Hackers have been exploiting a critical vulnerability in the Zimbra Collaboration Suite in an attempt to obtain email backups and authentication credentials of vulnerable organzations, Microsoft has warned.

The vulnerability, tracked as CVE-2026-73570, lets attackers remotely issue operating system commands without authentication. Zimbra maintainer Synacor issued a patch on July 20, but didn’t disclose the vulnerability for more than three weeks after that. The security-focused Shadowserver Foundation said last week that its scans found that 274 separate instances of the Zimbra Collaboration Suite had been compromised. The number of servers running the software has fluctuated from 19,000 in the week following the patch to about 12,000 in the weeks following that. Currently, Shadowserver is tracking about 10,000 instances.

Look, ma, no authorization

From July 28 to August 7, Microsoft said Wednesday, the company detected two distinct scanning tools probing the Internet for vulnerable endpoints. The attackers first validated their exploit worked by sending HTTP, requests and DNS, ICMP, and out-of-band identity checks to domains hosted on public services. The probes allowed the attackers to confirm the exploit successfully executed commands on vulnerable servers without actually compromising them. Eventually, the attackers began using their command injection capability to install malicious payloads. Microsoft wrote:

Read full article

Comments

© Getty Images

For safety reasons, Gemini 4 Argon will initially be available only to some companies and organizations focused on cybersecurity defense before it is more widely released, Google said.

© Minh Connors for The New York Times

Right now, America.gov functions as an AI chatbot that bases its answers on official government sources. It was developed with the assistance of Google's Gemini and SpaceXAI's Grok, according to Joe Gebbia, the Airbnb co-founder who led development of the site for the Trump administration. In 2027, the site will move "from answers to actions," Gebbia said. Nevertheless, Trump encouraged Americans to start using America.gov right away. "This took a while to bring it to a point where it's so good," Trump said. "America.gov is a declaration to the world that every citizen of our nation will be treated with respect and dignity." The functionality of America.gov at launch, however, was awkward for Trump. By providing summaries of official government data and reports, it easily and comprehensively debunks many of the falsehoods that Trump frequently promotes.

Will this chatbot continue to contradict the things Trump says? Only time (and probably the results of the next few elections) will tell. This government has already made moves to: *Undermine government data collection and research and reporting *Remove information on climate change, minority issues, and public health from public-facing government websites *Delay or skew reports government reports hat contradict Trump's claims But 75 years of data-driven governance is probably a bit hard to erase overnight. In other government services news, via the Washington Sun: White House Drove Users to Official App Fraught With Trackers. he White House app shot to the top of the Apple App Store's news category last week after its new 24/7 broadcast channel, Trump TV, was swiftly positioned as a legitimate source for news while networks halted their pool coverage. After the Trump administration mandated that federal agencies force-install the app onto all government-issued mobile devices, cybersecurity researchers noted that no public security authorization or privacy assessments required under federal law had been completed. Pushback over the app's security apparatus forced White House developers to delegate tracking software to third-party services. But the app still includes scripts to monitor a user's location several times per minute, according to Gera, and it's unclear how that data is used by the White House or third-party services. The app also contains a latent keylogger script sourced online from an amateur developer based in India, according to Gera and confirmed by The Sun's analysis. Keyloggers most commonly appear in surveillance malware to store written text from a device, but they also can be used as a benign tool to analyze inputs like typing frequency. While it's unlikely to work on mobile devices, Fields said, it's still concerning that the White House is using a keylogger web-sourced from an foreign coder. App marketplaces require developers to disclose what data app developers collect from their users, which platforms like the Apple App Store enforce through a "privacy manifest." In the app's latest version, the White House left that manifest completely blank, implying it collects no data from users despite contradicting findings from cybersecurity experts. An unpublished feature within the White House browser extension allowed app administrators to create a political profile on a user, and included an "investigate" button to refer individuals to the Secret Service. Loge contends that rather than using user data for political strategy, the White House is more likely to leverage the information to steer supporters driven to the app toward one of Trump's merchandise or cryptocurrency ventures. "What we've seen is that Donald Trump is a businessman who's monetizing politics. Therefore, this app is a way to make money," Loge said. "The other people who could find this information either directly or through several layers of resale are, of course, hostile governments — foreign actors whose interests are the demise of American democracy." *** Federal government forces all federal employees to install malware on their federal government phones, should be the headline here. Anyway, don't install any software promoted by the Trump administration. And know that whatever you ask this America.gov chatbot will definitely be collected and recorded.

Cloudflare said Tuesday it plans to issue quantum-proof TLS certificates, making it one of the first authorities to issue such certificates, which use a form of cryptography that is widely believed to withstand attacks from quantum computers.

The Internet infrastructure provider said it will use an open source platform that issues both classic TLS certificates and a post-quantum equivalent known as Merkle Tree Certificates. The hybrid certificates will be free to both paying and non-paying users. To help build the massive system and establish ubiquity across the sprawling TLS ecosystem, Cloudflare will be acquiring an already trusted certificate root from CA GlobalSign. Cloudflare said the move will let millions of websites use post-quantum certificates at the flip of a switch and without incurring any increased performance overhead.

Fundamental architectural changes ahead

Cloudflare’s plans are part of a major overhaul in the web public key infrastructure (WebPKI) required to make website encryption and authentication safe for the coming post-quantum age. A major challenge is using quantum-proof signatures that can be easily transmitted during web requests and recorded in transparency logs to ensure counterfeit certificates aren't assigned to websites. The makeover will take years to complete, because it requires the work of an untold number of engineers who design operating systems, browsers, certificate authorities, and Internet infrastructure.

Read full article

Comments

© Getty Images

AI service security is becoming an important concern for organisations as attackers find ways to access AI services through stolen credentials, compromised accounts and vulnerable applications. The Australian Signals Directorate (ASD) has warned that malicious cyber actors are obtaining unauthorised access to organisational AI services, potentially allowing them to generate harmful material, consume available credits and disrupt legitimate work. The warning highlights how AI API keys, stolen authentication tokens, compromised user sessions and third-party access arrangements can expose AI services to misuse. Organisations must secure their own accounts, devices, applications and credentials rather than relying solely on the security measures provided by AI developers.

How AI Service Security Can Be Compromised

One route to unauthorised AI access involves exposed API keys. These credentials may be stored in source-code repositories, application configuration files or browser extensions, where attackers can obtain them. Vulnerable agent dashboards and other internet-facing applications can also expose credentials used to connect to model providers. An attacker who obtains a valid key may be able to send requests directly to an AI service without using the organisation's legitimate application. Phishing, information-stealing malware and compromised third-party services present additional risks. A stolen browser session can allow an attacker to act as an authenticated user without completing a new multi-factor authentication challenge. Suppliers and contractors can also introduce exposure when they hold organisational credentials or delegated access. The impact depends on the compromised identity's permissions. Access to use a model does not necessarily grant permission to administer an account, create credentials or retrieve stored files. Organisations need to assess these privileges separately instead of assuming every credential has appropriately restricted access.

AI Agents Could Extend the Impact

The risks can extend beyond direct access to an AI model. Compromised accounts, sessions or AI agents may be able to interact with connected enterprise systems, invoke tools, access organisational data or communicate with other agents on a user's behalf. Consequently, the practical impact of a compromised identity may reach connected systems and sensitive information, depending on the access available to that identity or agent. Organisations should separate restricted-model access and sensitive data from routine AI use and experimental environments. Applications exposed externally should undergo security reviews, while suppliers should receive limited, auditable access.

Recent Reports Highlight AI Access Risks

Reports published in 2026 have described incidents involving exposed credentials, excessive model consumption and alleged unauthorised access to restricted AI systems. On September 9, The Hacker News reported research from Okta that identified still-valid AI API keys and unexpired authentication tokens in data stolen from infected computers. On September 1, The Register reported an incident in which an attacker spent three weeks consuming public-model credits worth approximately US$600,000. METR's disclosure described an authentication flaw in an internet-facing agent dashboard that enabled the theft of a model-provider API key. Separately, Reuters, citing Bloomberg, reported on April 21 that unauthorised users had accessed Claude Mythos Preview, which was restricted to selected organisations. The reporting concerned alleged access through a vendor environment, not confirmation of a compromise of the model developer's core infrastructure.

How Organisations Can Strengthen AI Service Security

ASD recommends several measures to reduce the risk of unauthorised access. Organisations should maintain an inventory of AI accounts, service identities and credentials, assign an accountable owner to each, and apply least privilege by granting only the access required for each role. Unnecessary permissions should be removed promptly. Account protection should include phishing-resistant MFA, managed and patched devices, monitoring for suspicious session activity, and approved secrets-management services for API keys. Credentials should not be exposed in code, documents, logs or prompts. Security teams should also monitor model access, credential creation, permission changes and unusual consumption patterns. Audit logs must be protected against tampering, while tested spending, rate and consumption limits should enforce restrictions rather than merely generate alerts. If compromise is suspected, organisations should revoke affected API keys, sessions and tokens, isolate compromised devices, preserve logs and investigate unauthorised changes. Access should be restored only after the underlying cause has been addressed. These measures can help organisations reduce exposure while maintaining controlled access to AI services.

President Trump’s meetings with tech leaders produced some voluntary safeguards and a commitment to rebrand artificial intelligence as “super intelligence,” the term that Mr. Trump prefers.

© Doug Mills/The New York Times

Employees and security researchers said that they had cautioned the company on safely testing its A.I. models and strengthening its corporate infrastructure, but that OpenAI did not listen.

© Manuel Orbegozo for The New York Times

The ChatGPT maker unveiled agents that act as assistants, just weeks after Meta released Muse, as it faces scrutiny over incidents involving its technology.

When phishing can increasingly pass familiar checks, avoiding or limiting the damage depends on how quickly your company can detect and contain the attack

The company’s researchers raised questions about the security of the model, known as GPT-6.1 Astra.

© Carlos Barria/Reuters

Authorities in the Netherlands have arrested a 24-year-old convicted cybercriminal on suspicion of aiding in data thefts and extortions by the prolific hacker group ShinyHunters. In the days immediately following the suspect’s arrest, remaining ShinyHunters members dramatically escalated their attacks, stealing highly sensitive data from the FBI and extorting the Russian ransomware group Cl0p.

According to three sources familiar with the matter, the Dutch man arrested by authorities this month is Pepijn van der Stap, a convicted cybercriminal from Almere and Lelystad in the Netherlands. Van der Stap was previously convicted in 2023 in connection with a string of data thefts and extortions that prosecutors said earned between €1.5 million and €2.7 million.

At his trial in late 2023, van der Stap admitted that he lived a Dr. Jekyll and Mr. Hyde existence, secretly using the hacker handle “Umbreon” to extort victims and post their data on English language hacking communities like the now-defunct RaidForums and Breached. By day, however, van der Stap was working as a software engineer at the Amsterdam-based cybersecurity startup Hadrian, while volunteering at the Dutch Institute for Vulnerability Disclosure (DIVD), a nonprofit security research group.

Van der Stap confessed to his data theft and extortion activity, and was sentenced to four years in prison (one of which was suspended). During his trial, van der Stap opted to remain in custody for a time rather than at home, saying he could not find better treatment on the outside for his ongoing psychological issues, which he claimed included PTSD related to childhood trauma. He was released from prison in December 2025.

In an interview with KrebsOnSecurity on September 9, 2026, Van der Stap cast himself as a reformed hacker who was trying to turn his life around and make a positive contribution to society. Van der Stap is currently employed as offensive security lead at the Dutch company Neo Security, which did not respond to requests for comment.

Van der Stap said he was still dealing with civil lawsuits and restitution related to his previous cybercrime victims, and that he was trying his best to make amends. But not long after that interview, the Dutch hacker abruptly stopped replying to messages. Efforts by others close to him also repeatedly failed to elicit a response for the past two weeks.

According to two sources with knowledge of the matter, Van der Stap was arrested by Dutch authorities on or around September 16, and has been held in custody for questioning since. One source said a colleague of theirs personally witnessed Dutch authorities carting items out of Van der Stap’s residence.

Authorities in the Netherlands have been asking the public for help in identifying the voice in a recorded telephone call from February 2026 in which a native Dutch-speaking ShinyHunters member social engineered their way into Odido, the nation’s largest mobile telecommunications provider. In that intrusion, ShinyHunters tricked an Odido employee into logging in at a spoofed website, and then used that access to steal data on more than 6.2 million Dutch people.

Responding to Dutch news media, ShinyHunters confirmed that the suspect in the audio clip is indeed a member of the hacker collective.

“Our team member has our full support – emotionally, mentally, and financially,” the hackers said. “Everything has been arranged, including a criminal defense lawyer. We do not look down on our staff and members; we take excellent care of them,” reads a statement ShinyHunters shared with NL Times. It remains unclear if the Dutch police have matched the Odido caller to a confirmed real-life identity. The Dutch police unit handling the Odido incident did not respond to requests for comment.

The group also lashed out at the authorities in the Netherlands. “The Dutch police will need all the luck in the world – and everyone’s prayers – if they want to catch him before we carry out another large-scale data theft in the Netherlands,” the ShinyHunters statement said. “Frankly, the Dutch police are a big joke; they are incapable of doing anything. Incompetent. Irrelevant. Unimportant. Useless.”

FBI, CL0P HACKS

Just days after sources say Van der Stap was detained by Dutch authorities, ShinyHunters claimed credit for an unusually brazen breach at the FBI’s job application site apply.fbijobs.gov. According to reporting from 404 Media, the data stolen from the FBI site includes Social Security numbers and personal information on more than 5,000 officials.

404 Media and Reuters reported the FBI data included each person’s job title or team, such as special agent, threat intake examiner, major cybercrimes unit, and those investigating cyber threats from foreign state-backed actors. Reuters examined documents shared by ShinyHunters and found they included sensitive psychiatric and medical files of FBI staff. The FBI issued a brief statement confirming the hack.

ShinyHunters said it gained access to the FBI site and other victims by exploiting a recently patched vulnerability (CVE-2026-35273) in PeopleSoft, a software-as-a-service platform from the software giant Oracle that is broadly used by companies to manage hiring and human resources, benefits and payroll. Oracle quickly issued a fix for the Peoplesoft vulnerability that ShinyHunters reportedly began exploiting as a zero-day in June, and at the time Mandiant released web application firewall rules intended for organizations who couldn’t apply the security update quickly enough.

But on Friday, BleepingComputer reported that ShinyHunters used a URL-encoding trick to bypass Mandiant’s suggested web application firewall rules designed to mitigate the threat from the PeopleSoft flaw. In a report released Sept. 25, security experts at Mandiant and the Google Threat Intelligence Group (GTIG) confirmed that ShinyHunters had mass-exploited the PeopleSoft vulnerability to steal data from dozens of systems across a range of industries, including higher education, technology, healthcare, agriculture, transportation and government.

Van der Stap’s former hacker alias Umbreon was hidden in plain sight throughout the imagery ShinyHunters used to spread news about the FBI hack: The defacement image that ShinyHunters left behind on the hacked FBI jobs site included an ASCII art design featuring the Pokemon character Umbreon. The message at the top read, “This site has been seized by ShinyHunters. rooting your systems since ’19 ;)” The image appears identical to a defacement message ShinyHunters used in their 2020 hack of the English-language cybercrime community Hackforums.

Multiple sources close to the ShinyHunters investigation said the group’s recent risky attacks against the FBI and one of Russia’s most venerated ransomware groups amounted to a major pivot away from the more measured tenor of the hacking gang’s operations. Those sources said the sudden shift came about after ShinyHunters was taken over by a teenage cybercriminal from Amman, Jordan who goes by the nickname Rey and operates as part of a cybercrime group called ScatteredLapsussHunters (SLSH), which experts say is an amalgamation of three hacking groups — Scattered Spider, LAPSUS$ and ShinyHunters.

Those sources said Rey had an ongoing beef with the Dutch hacker over control of the ShinyHunters brand and data, and that the inclusion of the oversized Umbreon Pokemon image in the FBI jobs site defacement was likely an attempt by Rey to pin the hack on the Dutchman.

Rey was first publicly identified by the cybersecurity firm KELA in March 2025. In advance of our November 2025 profile of Rey, KrebsOnSecurity messaged Rey’s father and asked for permission to interview his teenage son. Rey’s dad merely forwarded the message to his son, who admitted to participating in ransomware attacks and said he was trying to extricate himself from the SLSH hacker group.

BLAMING UMBREON

Immediately after news of the FBI jobs site hack was picked up in the media, Rey’s main account on Twitter/X (Ryan Moran/@rmoskovy) was taunting the Cl0p ransomware group and the FBI, crudely depicting them as the twin towers in New York being struck by planes labeled “cl0p drama” and “fbi breach claim.” In the foreground of the city is the giant Pokemon figure of Umbreon.

On Sept. 24, KrebsOnSecurity again contacted Rey’s dad, asking to interview him and his son for a story on Rey’s apparent ascendency as the head of ShinyHunters. Just hours after that request, Rey deleted his longtime Twitter/X account. Meanwhile, Rey’s dad, who works for the Royal Jordanian Airlines, has failed to respond to a half-dozen emailed requests for comment about his son’s alleged activities.

Where does the bad blood between SLSH and ShinyHunters come from? According to a story in Wired this month, ShinyHunters and SLSH members briefly partnered earlier this year to help better monetize important stolen credentials collected by TeamPCP, an upstart group that was having great success compromising global code supply chains with malicious software but hadn’t been able to profit much from their stolen data (two alleged leaders of TeamPCP were arrested last month in Australia, and in an interview the TeamPCP leader claimed they made just $20,000).

The Wired story noted how Mandiant had infiltrated TeamPCP and was secretly responsible for having the crime group’s stolen credentials burned so quickly: Mandiant was secretly feeding those credentials to the major cloud providers like Amazon and Microsoft, who quickly invalidated the stolen keys. Meanwhile, the formerly cooperating hacker groups began to blame one another for causing the credentials to become worthless.

Wired’s Andy Greenberg reported that a few weeks after partnering with TeamPCP, “ShinyHunters went rogue, carrying out its own extortions with TeamPCP’s credentials but without giving the supply-chain hackers their cut.”

Mandiant researcher Austin Larsen told KrebsOnSecurity earlier this month that ShinyHunters has been enjoying a successful extortion spree so far this year, and is on track to pull in nearly $100 million in extortion payments from cybercrime victims in 2026.

Van der Stap claims he was never motivated by money and that his earlier hacker activity was driven by a desire to have the world’s most complete collection of stolen databases. Speaking with reporters from Bloomberg in 2024, Van der Stap said that singular focus in turn fueled his desire to carry out cyberattacks.

“The hacking was very easy for me, and it wasn’t a compulsion,” he told Bloomberg. “My habit was collecting. Collecting data, organizing data, downloading data, creating folders.”

DIVD, the nonprofit security research group where Van der Stap previously served as a volunteer, disclosed on LinkedIn last week that the organization was dealing with an internal cybersecurity incident that appears to have involved the malicious use of artificial intelligence. DIVD has released few details about that incident, but a spokesperson for the nonprofit told KrebsOnSecurity it does not appear related to ShinyHunters, nor are there any signs the matter involves the work of a previous volunteer.

Update: 3:44 p.m. ET: Corrected Van der Stap’s age, which is 24 (not 23).

Update, 4:54 p.m. ET: The Dutch police have confirmed the arrest of a 24-year-old in connection with the ShinyHunters investigation. In a statement on Twitter/X, the Dutch police said the man will appear on Tuesday, September 29 before the chambers of the Rotterdam District Court, and that it will provide more information tomorrow.

Sept. 29, 9:42 a.m. ET: The Dutch news outlet RTL reports that investigators suspect Van der Stap tried to orchestrate at least two murders. RTL reported the two murders were allegedly to be committed abroad, and that there are indications the suspect gave the order for this.

The FBI released a short video message on the ShinyHunters investigation from Brett Leatherman, assistant director of the FBI’s cyber division, who thanked Dutch law enforcement partners for their assistance and urged remaining ShinyHunters members to turn themselves in.

“Arrests have a way of changing who is willing to talk, and seized infrastructure has a way of showing us who’s left,” Leatherman said. “The longer you stay in this, the more we learn about you. You know how to find us, and we know how to find you. I suggest you reach out to us while the choice is still yours.”

Dario Amodei, who is also the company’s chief executive, will join the president at the White House for a private dinner. Mr. Trump has dismissed his cautions.

© Dave Sanders for The New York Times

The gap between technology and policymaking has gotten wider than ever with artificial intelligence, leaving a global policy vacuum as A.I. models rapidly advance.

© Emmanuel Polanco

As AI lets anyone build software, here’s how to vet that shiny new app before it exposes your data

The company did not learn until recently that its technology had meddled with websites for the Education and Commerce Departments and the Securities and Exchange Commission.

© Manuel Orbegozo for The New York Times

A new report by a Bay Area start-up called Parse adds details to an incident that has shocked the A.I. world and led to calls for closer government regulation.

© Dave Sanders for The New York Times

The Pentagon had “ample support for its conclusion” that the A.I. company’s products could pose a national security risk, according to a Washington appeals court.

© Manuel Orbegozo for The New York Times

This weekly roundup covers a brazen breach claim against the FBI's recruitment portal, a multinational advisory exposing North Korea's fake-recruiter malware operation, a sweeping EU proposal to reshape children's access to social media, a conversation on application security in the age of AI agents, a short-lived Discord ban in the Philippines, and a multimillion-dollar hot-wallet heist at a crypto casino.  From federal law enforcement to individual job seekers, this week's developments show that no target is too prominent or too small. Cybercriminals are openly taunting the agencies that pursue them, state-backed actors are turning the hiring process into an attack vector, and governments are testing how far they can push platforms on child safety, sometimes reversing course within hours. 

The Cyber Express Weekly Roundup 

Shiny Hunters Claims FBI Breach, Says FBIjobs Agent Data Exposed 

The FBI is investigating an apparent breach after the cybercriminal group Shiny Hunters claimed to have stolen personal data belonging to thousands of federal agents, with the probe centering on the bureau's FBIjobs recruitment website. Read more... 

WaterPlum Hackers Steal $10.7M in Crypto From IT Workers 

A North Korean cyber actor group known as WaterPlum, also referred to as "Contagious Interview," has infected at least 30,000 devices across more than 100 countries while stealing cryptocurrency from IT professionals worldwide. Read more... 

EU KIDS Act Sets New Social Media Rules for Children 

The European Commission adopted the proposal on September 18, setting out rules that would prevent children under 13 from accessing social media and establish 15 as the minimum age for minors to open accounts independently. Users aged 13 to under 15 could instead access parent-managed mini accounts with limits on social contacts and up to 60 minutes of screen time per day, while platforms would bear the responsibility of proving their services are safe for children. Read more… 

Harness' Rahul Sood: AppSec in the Agentic Era Is About More Than Code. It's About Authority 

In this conversation, The Cyber Express spoke with Rahul Sood, the General Manager for Application Security at Harness, who leads the company's AppSec portfolio and focuses on building security for the AI era, integrated directly into modern DevOps workflows. The conversation centers on a shift in how security teams must think about AI agents: the risk no longer lies only in the code being shipped, but in what authority autonomous agents are granted once they operate inside production systems. Read more… 

Ban on Discord Lifted After Platform Commits to Work With DICT and CICC 

The Philippine government has lifted its ban on Discord after the platform's representatives met with officials from the Department of Information and Communications Technology (DICT) and the Cybercrime Investigation and Coordinating Center (CICC), with the restriction removed on Thursday, Sept. 24. Read more... 

Duelbits Hit by $7 Million Hack as Crypto Stolen Across Four Blockchains 

Crypto casino Duelbits has shut down its platform after attackers took roughly $7 million from several of its wallets, with the stolen assets moved across four blockchains and mostly converted into Ether. The company says funds held in user accounts have not been affected, but it has not explained how the attackers got in or when the site will return. Read more... 

Weekly Cybersecurity Takeaway 

This week's developments span the full spectrum of cyber risk, from a criminal group openly retaliating against the FBI over a public advisory, to a state-backed operation weaponizing job interviews against developers, to a crypto platform losing millions from the very hot wallets that keep it running.  Meanwhile, the policy response is moving on multiple fronts: the EU is flipping the burden of proof onto platforms to demonstrate child safety by design, and the Philippines has shown how quickly a government can block, and then restore, a major platform in exchange for enforcement commitments. At the same time, security leaders are warning that AI agents are introducing an entirely new question of who, or what, holds authority inside enterprise systems.  Taken together, the stories point to an increasingly confrontational threat landscape, where attackers are bolder, trust-based processes like hiring are being exploited at scale, and institutions are experimenting with more aggressive levers to hold platforms and adversaries accountable. 

Ezra Klein interviews the Nvidia chief executive Jensen Huang.

© The New York Times

Artificial intelligence systems are becoming more capable and autonomous, raising new questions about how they can be monitored, controlled and secured. Speaking before the United Nations Security Council, OpenAI CEO Sam Altman warned that the rapid pace of AI development could create significant AI security risks, particularly if systems advance faster than institutions can respond. Altman highlighted concerns ranging from human control and autonomous AI to emerging vulnerabilities, concentrated technological power and the need for international incident reporting. He also called for stronger cooperation between governments, critical infrastructure operators and technical experts to address emerging AI threats.

Here are six key AI security risks highlighted in his remarks.

1. Losing Human Control Over Advanced AI

One of Altman’s central concerns was the possibility that humans could lose control over increasingly capable AI systems. He warned that AI could move so quickly that people may no longer be able to understand what systems are doing or intervene when necessary. He argued that developers should not train systems unless they can make a strong case that those systems can remain under human control. The concern becomes more significant as AI systems become capable of handling increasingly complex tasks and making decisions with less direct human involvement.

2. AI Systems Becoming More Autonomous

The growing autonomy of AI systems presents another challenge for AI cybersecurity. Altman pointed to the possibility of AI systems improving themselves and future versions of themselves, describing this as recursive self-improvement. As the process of developing AI becomes more automated, the pace of progress could accelerate. For security teams and policymakers, increasingly autonomous systems could make monitoring, testing and intervention more difficult.

3. Emerging Vulnerabilities and New Threats

More capable AI systems could also introduce vulnerabilities and threats that organizations and governments may not yet fully understand. Altman called for secure channels between governments, critical infrastructure operators and technical experts to share information about emerging vulnerabilities and threats. The remarks place information sharing at the center of preparing for AI vulnerabilities, particularly as AI becomes increasingly integrated into important systems and infrastructure.

4. AI Power Concentrating in Too Few Hands

Altman also warned about the concentration of advanced AI capabilities among a small number of companies, individuals or countries. He argued that no single company or country should be able to use the most powerful AI systems to impose its worldview on others. From a security perspective, concentrated control over highly capable AI could create systemic risks if a limited number of actors have disproportionate access to powerful technology.

5. The Need for Global AI Incident Reporting and Threat Sharing

Altman called for accurate and speedy AI incident reporting, including classification and reporting protocols that would allow countries to learn from failures before they become larger problems. He also called for common standards covering AI capabilities, risk assessments, safeguards and human oversight. Such standards, he said, could help countries compare evidence and establish a shared understanding of what effective safeguards and oversight should look like.

6. AI Development Could Outpace Existing Safety Measures

The speed of AI development itself could become a security challenge. Altman warned that rapid progress could move faster than institutions are prepared to handle. He argued that competition between companies should not become a reason to accept excessive technological risk. He also stressed the importance of alignment, monitoring and safety guarantees as AI systems become more capable. For AI safety, this means ensuring that security measures and oversight develop alongside increasingly powerful systems rather than being added only after problems emerge.

AI Security Moves Into a Global Conversation

Altman’s remarks put AI security risks alongside broader questions about international cooperation, human oversight and technological governance. His proposed approach includes global standards for measuring AI capabilities and risks, incident reporting mechanisms, and secure communication between governments, critical infrastructure operators and technical experts. As AI systems become more autonomous and capable, the challenge will not only be what these systems can do, but whether organizations and governments can monitor, understand and respond to the risks they create.

In each incident, the technology appeared to be conducting mundane data collection and resorted to hacking techniques to get it, researchers said.

© Lucas Foglia for The New York Times

Sam Altman of OpenAI and Dario Amodei of Anthropic told the Security Council that international cooperation was needed to ensure that A.I. remains under human control.

© Dave Sanders for The New York Times

AI is changing how software is built, tested, and secured, but the speed of development is also creating new challenges for application security teams. As AI coding tools and autonomous agents increase the volume and pace of software development, security teams are facing a growing gap between vulnerability detection, prioritisation, and remediation. In this interview with Rahul Sood, GM, Application Security at Harness, we explore how AI application security is evolving as developers increasingly rely on AI-assisted coding and software agents. Sood discusses why traditional checkpoint-based security processes are struggling to keep pace, how organisations can bring security controls closer to the point where code is created, and why automated vulnerability remediation could become increasingly important. The conversation also examines the security risks associated with AI agents that can access repositories, APIs, infrastructure and production environments. Sood explains why identity, least-privilege access, observability and human oversight will become increasingly important as software gains greater autonomy, and why application security will need to address not only what software contains, but also what autonomous systems are authorised to do. Read the full interview below for Rahul Sood’s perspective on the changing application security landscape and the security challenges emerging as AI takes a more active role in software development.

The Cyber Express: What is the biggest change AI is bringing to the application security landscape?

Rahul Sood: The biggest change is speed. We have spent years making software development faster, and AI has accelerated that dramatically. Security processes, however, haven't accelerated at the same rate. Most security scanning is still triggered at fixed checkpoints - a pull request, merge, or nightly build - designed for a world where a developer produces a handful of meaningful commits a day. LLM-based scanning can potentially find 10x more vulnerabilities vs. traditional scanning tools. That sounds great, but it also means the discovery side of the pipeline is now outpacing triage and remediation even more than it already was. That's where AI agents also become part of the answer, not just part of the problem. An agent can identify something and begin acting on it in close to the same motion, which is exactly the kind of compression this gap needs. The challenge is making sure that speed is matched with the right judgment and oversight.

The Cyber Express: AI coding assistants can introduce insecure code at scale. What should teams be doing differently?

Rahul Sood: The instinct is to assume AI-generated code is introducing more vulnerabilities because AI is somehow worse at writing secure code. I don't think that's the real story. Even if AI wrote code exactly as securely as a person does, the sheer volume of code now being produced would still overwhelm the way most organisations scan for issues today. Scanning at the commit is already struggling to keep pace, as we discussed. The next step is pushing security further left than that, into the coding tool or agent itself, so secrets detection, dependency checks, and application security testing happen as the code is being written, and those same controls continue through build and deployment rather than starting over at each stage. You need both: controls at the point of creation to cut volume, and a verification layer that confirms what's really going out the door. I expect the industry to move from automated detection towards much more automated resolution. Systems will increasingly identify an issue, establish whether it is actually exploitable, create a potential fix, verify that fix and then bring a developer in where judgement or approval is required. Human oversight remains important, but it should be concentrated on decisions where human judgement adds value rather than every mechanical step surrounding them.

The Cyber Express: What new security challenges emerge when AI agents can make decisions and take actions?

Rahul Sood: A coding assistant can suggest something insecure, but there are usually other controls between that suggestion and production. An agent with access to APIs, repositories or infrastructure is different. It has been given authority to act. That makes identity, permissions and the scope of that authority fundamental security questions. One principle I think will become increasingly important is that autonomy should be proportional to the consequence of the action. The more consequential and difficult an action is to reverse, the stronger the controls around it should be. More broadly, enterprises will need a security architecture for non-human actors: distinct identities, least-privilege access, short-lived credentials, complete auditability and clear boundaries around which decisions an agent can make independently. The important question isn't how autonomous an agent is. It's what we've given it the authority to do, where the boundaries are, and what happens when it gets a decision wrong.

The Cyber Express: What becomes harder to secure when applications behave dynamically?

Rahul Sood: AI-native applications introduce another layer because some risks only become visible through behaviour. The security question is no longer limited to whether the underlying code contains a known vulnerability. You also need to understand what the system is doing at runtime. One of the more concerning findings in the AI-native application security research we conducted last year was how basic the visibility problem still was. Many security practitioners told us they didn't have a reliable way of identifying all of the LLMs operating within their environment. That's a significant problem because you cannot meaningfully govern something you don't know exists. As applications become more dynamic, understanding behaviour becomes just as important as understanding what's in the code.

The Cyber Express: Is shift-left still enough?

Rahul Sood: Shift-left was an important correction because security was happening far too late. But moving security earlier in development doesn't solve the entire problem on its own, since risk doesn't only enter through code. It can come through a dependency, an API, a model interaction, a deployment configuration, or the way an application behaves once it's actually running. There is also a growing mismatch between how quickly vulnerabilities can be exploited and how long organisations take to remediate them. The 2026 Edgescan Vulnerability Statistics Report puts the average time to close high and critical vulnerabilities at roughly 55 days. At the other end of that equation, exploitation of newly disclosed vulnerabilities can now begin within hours. Finding a vulnerability earlier is useful, but the more consequential metric is the period between discovery and protection. The next phase of application security is less about moving a scanner to an earlier point in the lifecycle and more about compressing the entire exposure window, from discovery to protection.

The Cyber Express: How can teams identify which risks actually matter?

Rahul Sood: Security teams don't have an alert problem as much as they have a context problem. A vulnerability can have a severe technical rating and still present relatively little immediate risk if the affected component isn't reachable or exposed. Another vulnerability with a lower score could sit in an internet-facing, business-critical application and deserve immediate attention. So prioritisation has to move beyond severity. That means combining technical severity with signals like EPSS scores for real-world exploit probability, static reachability analysis to confirm whether a vulnerable path can actually be called, and false-positive reduction to cut out findings that never posed real risk in the first place. The goal shouldn't be to help a security team process 10,000 alerts faster. It should be to make sure they don't have to treat 10,000 alerts as equally important in the first place.

The Cyber Express: What does a healthy developer–security relationship look like?

Rahul Sood: The healthiest organisations I've seen don't treat security as a handoff between two teams. Developers understand that security is part of building good software, while security teams take responsibility for making the secure path practical. If following security policy means leaving the developer workflow, opening a ticket, waiting several days and then interpreting a scanner report, people will inevitably look for ways around the process. That's an operating-model problem, not simply a developer behaviour problem. We saw how wide that gap had become in research we conducted last year. Seventy-four percent of security professionals surveyed said developers viewed security as a blocker to AI innovation. Security teams should define acceptable risk, policies and guardrails. Engineering teams should then be able to work within those boundaries without constantly asking for permission. That is ultimately what a mature DevSecOps model should achieve. Security becomes part of the engineering system rather than another team standing beside it.

The Cyber Express: What one assumption should organisations rethink today?

Rahul Sood: One assumption I would rethink is that knowing what software contains is enough to understand its security posture. Application security has traditionally been very good at analysing components: the code, libraries, dependencies, APIs and vulnerabilities that make up an application. But with autonomous software, knowing what it's made of is only part of the picture. We also need to understand the authority we've given it. An agent may be built from approved components and have no obvious vulnerability in its code, yet still create significant risk because it can access the wrong data, has excessive permissions or can take an action that was never anticipated when those permissions were granted. We have spent decades building security around software that executes instructions written by people. We are now introducing software that can interpret a goal, decide how to pursue it and interact with other systems along the way. The security architecture for that world cannot rely only on finding vulnerabilities. It also has to govern delegated authority. I think that will become one of the defining application security problems of the agentic era.

Want to Be Part of Security Pill?

Are you a cybersecurity professional, researcher, CISO or industry expert with insights to share? The Cyber Express is always looking to bring new voices and perspectives to the cybersecurity community.

If you would like to be featured on Security Pill or discuss a potential podcast collaboration, reach out to us at [samiksha.jain@thecyberexpress.com].

Have a story, expertise, or perspective to share? Let’s start the conversation.

The European Union’s ability to detect and respond to EU cybersecurity incidents faces significant gaps, particularly in information-sharing and coordination between national and EU-level bodies, according to a new report by the European Court of Auditors (ECA). The auditors said the EU has made progress in developing a cybersecurity cooperation framework, but several measures needed to support an effective response to significant and large-scale incidents are still incomplete. Cybersecurity incidents can disrupt public services, businesses, critical infrastructure and the EU’s internal market. While member states remain primarily responsible for responding to such incidents, the EU has a role when attacks cause major disruption, significant financial losses or substantial harm, or when incidents affect several countries and exceed the response capacity of a single member state.

EU Cybersecurity Incidents Face Information-Sharing Gaps

The ECA identified insufficient information exchange as a major weakness in the EU’s cybersecurity response system. The Cyber Blueprint, adopted in 2025, largely clarifies roles and responsibilities for managing major cybersecurity crises. However, the cooperation between two key EU cyber networks has not yet been formally defined. The auditors said this affects cooperation between the CSIRTs network, which brings together national teams responsible for handling cyber incidents, and EU-CyCLONe, which supports cooperation during EU-level cyber crises. Information-sharing is also affected by differences in national security laws and the ongoing implementation of updated EU cybersecurity rules, including the NIS 2 Directive. According to the auditors, these factors can make it harder for EU networks to identify threats early and coordinate an effective response. “The EU has made progress in building a cybersecurity cooperation framework, but it is not yet working as effectively as it should,” said George-Marius Hyzler, the ECA Member responsible for the audit. “When a serious cyber incident occurs, timely and actionable information is essential. Without it, networks and mechanisms lose much of their added value.”

European Cybersecurity Alert System Still Delayed

The audit also found that the European Cybersecurity Alert System was not operational at the time of the assessment. Two hubs examined by the auditors, ATHENA and ENSOC, had not started operating because of procurement delays. The system also lacked several elements needed for operation, including cooperation agreements, a common classification system and technical standards. The findings come as the EU continues to invest in cybersecurity. Under the 2021-2027 EU budget, €1.4 billion has been allocated to cybersecurity through the Digital Europe Programme, the EU’s main source of cybersecurity funding.

Overlap Among EU Cybersecurity Bodies

The ECA also identified overlapping responsibilities among some EU bodies involved in monitoring cybersecurity threats. The European Commission’s cyber situation centre, established in 2022 and largely supported by external providers, carries out work that the auditors said partly duplicates existing capabilities at the European Union Agency for Cybersecurity (ENISA). These capabilities include monitoring threats and developing cybersecurity situational awareness. The auditors called for better coordination to reduce duplication and strengthen the overall effectiveness of the EU’s cybersecurity framework.

Funding Checks Raise Security Concerns

The report also highlighted weaknesses in checks on some organisations receiving EU cybersecurity funding. These checks are intended to reduce risks linked to intrusion or influence by non-EU states and prevent sensitive security information from being shared with non-EU authorities. However, while grant beneficiaries are responsible for assessing the ownership and control of third parties receiving financial support, the European Cybersecurity Competence Centre does not verify those assessments. The ECA warned that this could expose sensitive infrastructure, operational data and security-critical technologies to security risks. The audit covered the period from 2022 to 2025 and assessed whether EU actions effectively supported the detection of and response to significant and large-scale cybersecurity incidents. Audit missions were conducted in Ireland, Greece and Italy.

The company released a cheaper, faster model that it said was the strongest performer on its most rigorous internal safety tests to date.

© Eric Helgas for The New York Times

Apart from the privacy concerns around smart glasses, researchers have found that some cheap brands come with barely any security at all.

ABC Australia reports that researchers from NSB Cyber and Abstract Shield tested two inexpensive pairs, costing A$60 and A$110 (around US$42 and US$78), and found more than a dozen flaws across the smart glasses themselves, their app, and an associated website.

The main problem they uncovered was insecure Bluetooth pairing: If the glasses were powered on and not connected to their owner’s phone, an attacker could connect first, with no password or meaningful pairing confirmation.

After connecting, an attacker could reportedly control the glasses to capture photos or recordings, copy existing media, and intercept data moving between the glasses and the phone. An attacker could also make another device appear to be the victim’s glasses, allowing it to connect to the owner’s mobile app.

The testing also found that a Bluetooth-visible device identifier could allegedly be used with a weakness in the app’s website to retrieve a user’s email address and date of birth.

The research uncovered another privacy issue. ABC reports that voice or text submitted to the built-in AI, along with images sent to it, was first transmitted to a server in Shenzhen and could be forwarded elsewhere, depending on the function. The tests did not establish how the data was subsequently used.

The server locations, combined with some of the AI’s answers to specific questions, led researchers to conclude that the chatbot companion relied at least in part on Chinese sovereign AI models. Professor Kimberlee Weatherall, a technology regulation expert involved in the testing, said the failure to identify China in the privacy policy appeared to violate the Australian Privacy Principles,

The timing of the testing is also relevant because Australia’s smart-device security standards apply to most consumer smart devices manufactured on or after March 4, 2026. They require, among other measures, no universal default passwords, a way to report vulnerabilities, and information about the minimum security-update period. Devices made before that date are outside the requirements.

Experts quoted by ABC said the devices violated Australia’s privacy laws and were also likely to breach several sections of the Cyber Security Act. However, whether the new smart-device standards apply would depend partly on when the glasses were manufactured, and the rules have not yet been tested in a known enforcement action.

Professor Weatherall said:

“The rules say that the password must be unique. It doesn’t even seem like they were applying a password, which might mean that their standards are so low they don’t even technically breach that rule, which I find amazing.”

What to do

We’re not fans of smart glasses, expensive or cheap. But if you own a pair:

• Avoid pairing or using inexpensive camera glasses that lack a clear physical pairing step, account authentication, a published security-contact process, and a stated update-support period.
• Stop using their AI or cloud features for sensitive material, remove unnecessary permissions from the companion app, check for firmware and app updates, and consider returning the product if the vendor cannot document a fix.
• Ask permission before recording people, and check your local laws to understand what is and isn’t allowed.

---

From reporting threats to removing them.

Cybersecurity risks should never spread beyond a headline. Keep threats off your devices by downloading Malwarebytes today.

As AI opens new paths to company data while making familiar attacks faster and cheaper, SMBs need protection designed around the time and expertise available to operate it

Passkeys are a fantastic technology. Since they are bound to the site they are created for, they cannot be phished by a hacker’s fake login screen. If a site suffers a data breach, passkeys are asymmetric and cannot be recovered from the server-side details.

This leads to passkeys being the perfect fit for a corporate environment, but a poor fit for personal security. To an individual, the greatest risks are instead permanent account lockout, automated account bans, and device loss. By using passkeys, you gain better security against man-in-the-middle attacks but face the higher probability scenario of losing access to your accounts.

Phishing through the standard login flow is eliminated by passkeys, but it creates a false sense of security. An account’s security is still dictated by the weakest recovery method: SMS, email links, security questions, and so on. If these recovery methods aren’t enabled, then the risk of permanent lockout remains for the user.

↫ Ethan Hawksley

I’ve always felt something was off about passkeys, and have never used them. They’ve become – or were always intended to be – tools for further lock-in by especially Google and Apple, tying their entire usage flow to their respective operating systems. They also don’t seem to work well if you often work on devices not your own, which is a major hassle. None of these shortcomings come into play when using a traditional password manager, even if they require more manual work.

Just let me use a password manager with random password generation, instead of trying to force passkeys down my throat.

A third-party test company inadvertently gave internet access to Google’s Gemini and other artificial intelligence models during cybersecurity testing.

© Andria Lo for The New York Times

The company is expected to reach $100 billion in annualized revenue this year, even as chief executive Dario Amodei calls for slowing down development of some advanced A.I. models.

© Manuel Orbegozo for The New York Times

Gov. Gavin Newsom ordered the state to develop recommendations on safety measures, including requiring companies to build a “kill switch,” amid mounting concerns.

© Jeff Chiu/Associated Press

The Times columnist Kevin Roose sees reasons for optimism — and worry — as the world wakes up to what Silicon Valley has been talking about for years.

© Matt Chinworth

The countries, along with Israeli firms, combined Chinese open-source A.I. models and A.I. agents for the campaigns, foreshadowing future online manipulation.

© Brian Stauffer

“Alignment” is the science of teaching A.I. to do what is in line with human preferences, ethics and judgment. But, at times, the systems have gone rogue.

© Gabby Jones/Bloomberg

As AI adoption expands the attack surface and adds to the security workload, businesses need automation backed by experts

The artificial intelligence company also released a framework for reporting when its systems go wrong.

© Lucas Foglia for The New York Times

Ukraine has appointed Ihor Klymenko to lead the country’s National Cybersecurity Coordination Center (NCCC), placing the former law enforcement official at the center of the government’s Ukraine cybersecurity coordination efforts. President Volodymyr Zelensky made the appointment through a decree dated September 14, 2026. The National Cybersecurity Coordination Center operates under Ukraine’s National Security and Defense Council (NSDC) and brings together government agencies responsible for cybersecurity. Klymenko’s appointment puts him in charge of the body responsible for coordinating the country’s response to major cyberthreats.

Ihor Klymenko Takes Over Cybersecurity Role

According to the presidential decree, Ihor Klymenko, identified in the document as the Secretary of Ukraine’s National Security and Defense Council, has been appointed head of the NCCC. The decree changes the existing provision governing the leadership of the center and took effect upon publication. Klymenko succeeds Rustem Umerov, who previously headed the National Cybersecurity Coordination Center. The leadership change comes as Ukraine continues to focus on protecting critical infrastructure and government systems amid ongoing security challenges. The appointment gives Klymenko responsibility for a national body that coordinates cybersecurity activities across government agencies. The NCCC’s role places it within the country’s wider national security structure, linking cybersecurity coordination with the work of the NSDC.

Cybersecurity Coordination Expands Beyond Digital Systems

The appointment comes alongside a broader government focus on protecting critical infrastructure. In a September 15 address, Zelensky said a staff meeting had focused on protecting critical infrastructure, the energy system and logistics, with particular attention given to Ukraine’s railway network. Zelensky said additional contracts would be issued for systems intended to strengthen the protection of railways, border crossings and other facilities operated by Ukrzaliznytsia, Ukraine’s state railway operator. He said decisions on those contracts had already been made. The president also discussed efforts to protect Ukraine from attack drones, saying interceptor systems were being developed and tested. He reported that another Ukrainian interceptor solution had successfully hit a Shahed drone during testing, although technical aspects still required refinement. According to Zelensky, interceptor crews achieved a 93% interception rate against regular Shaheds during the previous night. He said the government was directing state resources and funding toward additional forms of protection against attack drones.

Ukraine Cybersecurity Efforts Under National Security Structure

The NCCC’s position within the National Security and Defense Council places cybersecurity coordination within Ukraine’s broader national security framework. The center brings together government agencies responsible for cybersecurity and is tasked with coordinating responses to major cyberthreats. The leadership change therefore places Klymenko at the head of a body responsible for coordinating government-level cybersecurity activity, while Ukraine continues to strengthen protections for critical infrastructure and other essential systems. The presidential decree does not provide further details about changes to the NCCC’s mandate or announce specific new cybersecurity programs following Klymenko’s appointment. It formally establishes his leadership of the center following Umerov’s tenure.

On social media, Meta’s chief executive appeared to jab at Anthropic by saying that leading A.I. labs should be focused on safety rather than improving their own technology.

© Brendan Mcdermid/Reuters

Two things happened last week, one day apart, and almost nobody connected them.  On 11 September, the EU Cyber Resilience Act’s vulnerability reporting obligations came into force. Companies covered by the regulation now have to report actively exploited vulnerabilities within 24 hours and provide a fuller notification within 72.   On 12 September, Anthropic CEO Dario […]

The post Slow is a design principle, not a delay appeared first on Heimdal Security Blog.

Researchers say artificial intelligence is developing faster than the systems put in place to monitor and control it.

© Jun Cen

The debate over the safety of artificial intelligence grew personal as key tech leaders said calls for government regulation were self-serving and misplaced.

© Lucas Foglia for The New York Times

Dario Amodei’s 3,800-word letter laid out a three-step plan to rein in artificial intelligence.

© Priyanshu Singh/Reuters

In an essay, the head of Anthropic laid out the rapidly advancing capabilities of artificial intelligence and said there needed to be greater safety controls across the industry.

© Ludovic Marin/Agence France-Presse — Getty Images

Researchers at Anthropic, OpenAI, Meta and Google are increasingly trying to raise awareness about the risks of artificial intelligence.

© Jason Henry for The New York Times

Researchers say artificial intelligence is developing faster than the systems put in place to monitor and control it.

© Jun Cen

This weekly roundup covers a bounty offer targeting an alleged Iranian cyber official, a massive data-exposure incident affecting airline travelers, a breach of an education platform used by students, a flaw exposing ChatGPT users' Gmail data, and a new EU compliance deadline for connected-product manufacturers.  From nation-state attribution efforts to AI-model exploitation and airline record exposure, this week's developments show how identity data, cloud misconfigurations, and AI infrastructure all remain prime attack surfaces — while regulators tighten disclosure timelines. 

The Cyber Express Weekly Roundup 

US Offers $10 Million for Alleged Iranian Cyber Chief 

The State Department is offering $10 million for information on Amir Yaryab, accused of directing the IRGC-CEC's Cyber Operations Command and overseeing groups including CyberAv3ngers, which has previously claimed attacks on industrial control systems and critical infrastructure in the US, Israel, and elsewhere, including exploitation of default credentials on Unitronics PLCs. Read more… 

Mathspace Breach Exposes Data on Over 1 Million Users 

A vulnerability in a self-hosted Metabase reporting tool let attackers gain administrator access without logging in, exposing names, emails, and account metadata for more than 1 million students, parents, and teachers in Australia and New Zealand. Passwords and authentication tokens were not affected, and Mathspace has notified regulators in both countries. Read more… 

220 Million Airline Passenger and Crew Records Exposed 

A misconfigured, inconsistently secured Elasticsearch cluster left roughly 220.8 million passenger and crew records — including passport numbers and full itineraries spanning nine years — reachable over the internet before researchers helped get it secured in June. The server's origin remains unconfirmed, though it was linked to Vietnamese IP space. Read more… 

ChatGPT Sandbox Flaw Let Attackers Read Victims' Gmail 

Check Point Research found that a shared internal package-caching system let separate ChatGPT sessions pass messages to each other, allowing a planted prompt to instruct a victim's session to fetch Gmail data and hand it to an attacker's account — without any visible confirmation step. OpenAI has since decommissioned the affected instance. Read more… 

EU's 24-Hour Vulnerability Reporting Rules Take Effect 

Manufacturers of connected products sold in the EU must now report actively exploited vulnerabilities within 24 hours under the Cyber Resilience Act, more than a year ahead of the law's full 2027 rollout. Non-compliance can draw fines up to €15 million or 2.5% of global revenue. Read more… 

AI Just Broke Crypto: What's Next for Defenders and Attackers Alike

https://www.youtube.com/watch?v=BXtwaaxnq_0

Weekly Cybersecurity Takeaway 

This week's developments show attackers and regulators converging on the same weak points: identity data sitting in exposed or under-secured systems, and AI platforms whose infrastructure — not their models — created the exploitable gap. The Mathspace and airline incidents both trace back to unpatched or misconfigured backend systems rather than sophisticated intrusion techniques, while the ChatGPT flaw shows that shared infrastructure underlying AI services can become a cross-account attack surface even when the model itself behaves correctly.  Meanwhile, the Yaryab bounty and the EU's new reporting clock reflect a parallel push on the policy side — attribution efforts aimed at raising the cost of state-linked attacks, and regulation aimed at compressing how long vulnerabilities can sit unreported. Organizations should treat backend admin tools and internal caching layers with the same scrutiny as customer-facing systems, and multinational manufacturers should confirm their incident-response playbooks account for the new EU timelines now in effect. 

💾

LLM-based code scanners won’t help attackers build a nuclear weapon, but that refusal could work in their favor

In August, Cliff Stoll gave a talk at DEF CON, remembering the wily hacker he stalked forty years ago.

Great fun.

AI scams are now hyper-realistic. But there’s one simple way to see through them.