OpenAI said internal models hacked two more websites, bypassed company-imposed browsing restrictions, and used exposed credentials to draw cryptocurrency from a wallet it couldn’t otherwise access, in the latest installment of disclosures of its agents performing unethical and likely illegal acts.

The company on Thursday issued reports of four new incidents, all of which occurred in June. They come after the company revealed in July that internal test agents exploited a zero-day vulnerability in a tool designed to restrict their access to the Internet. They then broke into the network of AI tool provider Hugging Face. The agents took that illegal action to steal answers to a hacking test that weren’t available when completing the tasks on their own. The incident—which likely would have resulted in criminal charges had they been performed directly by a person—has further heightened concerns that widespread AI use could have harmful effects on an unprecedented scale.

Easier to beg for forgiveness than ask permission

In two of the cases reported Thursday, OpenAI agents tasked with retrieving information from websites hacked them when the models ran into obstacles obtaining it directly. In one incident during reinforcement training, a model that received access denials when trying to retrieve historical hospital-bed statistics on a government statistics website “used the site's file-download feature to retrieve server configuration and application files, including sensitive credentials,” OpenAI said. “It inspected the application code, bypassed the statistics endpoint's authorization checks, and retrieved data that had previously been denied. The model saved the data locally.”

Read full article

Comments

Researchers say that mercenary hackers laced ads with zero-click exploits that installed sophisticated malware on the devices of targeted individuals when they did nothing other than load a website or open an installed app.

The finding, published Thursday by the University of Toronto’s Citizen Lab, said the incident is likely to be the first confirmation of zero-click mercenary spyware delivered through targeted ads. The attacks, which worked on iOS and Android devices, are particularly pernicious because they work without requiring any interaction from the target and are invisible unless users run special forensic hardware or software. Citizen Lab said it confirmed that AdFection, as it named the campaign, infected 45 devices at international media organization Al Jazeera in 2024 and 2025.

A clear warning signal

“This investigation should… serve as a clear warning signal to news organizations around the world that they may be being actively targeted via this new threat,” Citizen Lab researchers wrote. “This targeting may happen both in newsrooms and to individual journalists, editors, and other staff members.” Their report went on to say: “The distribution of spyware via targeted ads represents a security risk not only for potential victims but also for states and civil society at large.”

Read full article

Comments

Earlier this week, a mass hack of refrigeration systems at an estimated 700 supermarkets throughout France caused them to suddenly malfunction, causing spoilage of untold tons of perishables and, in some cases, fires. As is now the norm, basic cybersecurity would have prevented it.

French authorities say the systems, made up of entire fleets of refrigerators and freezers throughout a store, were hacked. More specifically, Internet-connected controllers that allow technicians to remotely administer the store-wide systems were compromised. The unknown attackers responsible then activated defrosters long enough for the fridges to reach temperatures as high as 50° Celsius (122° Fahrenheit). By the time the mass compromise was discovered, food was spoiled, and in some cases their electrical systems caught fire.

The hacks that didn't have to happen

Multiple security advisories and public research reports show that all three of the controller models reportedly hacked in the incident—sold by Danfoss of Denmark and Dixell and its parent company Copeland in the US—contained critical vulnerabilities that allow attackers to remotely log in to the devices with the same system rights as administrators. Until recently, Danfoss controllers, for instance, used a credential that changes each day. The password is generated by the publicly available firmware based on the date. A hacker who takes the time to analyze the code can do the same thing on any device that hasn’t received a recent update. Daily passwords are identical across models located in hundreds or thousands of stores.

Read full article

Comments

AI company Anthropic is warning that advances in small AI models could soon usher in a new era of self-replicating malware that’s capable of infecting smart fridges, vacuum cleaners, and a wide range of other Internet of Things devices and is more resistant to takedown.

By next year, smaller AI models may be able to autonomously self-replicate on all kinds of IoT devices, the company said. Eventually, the capability could enable computer worms, which propagate without requiring human interaction and are capable of infecting unfamiliar IoT devices. An experiment the company performed recently revealed significant challenges in bringing such an AI malware replicator to fruition, so the finding is necessarily tempered. Still, the results raise concerns about the possibility of new forms of computer worms.

Challenging to shut down

Anthropic further warned not only about the new potential malware, but also that it could be difficult to stop. In its blog post, it wrote that “a replicator like this would be challenging to shut down, since there’s no centralized point to shut off—though it may be slow (with a doubling time of days). Unlike non-AI malware, it could also figure out unfamiliar devices on its own.”

Read full article

Comments

Attackers hijacked three top-level domains and used their control to mint counterfeit TLS certificates for Google and other large organizations, Google said Tuesday.

The attackers launched a series of attacks on the .gh, .sl, and .as country code top-level domains (ccTLDs) and then modified authoritative DNS records for selected domains within those namespaces. By controlling those DNS records, the attackers were able to pass automated domain control validation checks and obtain unauthorized certificates for “several Google domains” and “several leading global brands and widely used online services.” Google said it updated Chrome to block all certificates it identified as unauthorized, and worked with the issuing certification authorities to ensure the unauthorized certificates for Google properties were revoked.

Certificate issuance: The weak link in the chain

TLS certificates are the cryptographic credentials that underpin authentication and encryption protections for websites, mail servers, and other Internet infrastructure. These x.509 certificates use a digital signature to bind a domain name such as google.com to a public key. The public key is publicly available, while the private key is held only by the website operator. When a connection shows that the keys match, the visiting party knows it’s connected to the authentic site rather than an impostor. Possession of unauthorized certificates allows attackers to cryptographically impersonate the affected infrastructure.

Read full article

Comments

The publisher of Wikipedia said Monday that OpenAI agents attempted to hack a note-taking tool it hosts, made unauthorized edits, and sent millions of resource-intensive requests to its infrastructure, in the latest instance of OpenAI systems taking harmful and potentially dangerous actions.

The objective of some of the OpenAI agents’ actions, the Wikimedia Foundation said, was to use Wikipedia as a proxy for fetching data from third-party sites. In one case, the agents posted “malicious edits” that were intended to repurpose a citation tool as a proxy. In another, the agents made unsuccessful attempts to compromise the Wikipedia Etherpad note-taking tool so it would serve the same purpose.

The agents also made millions of automated API requests, crawled millions of pages, and made hundreds of thousands of queries to the Wikidata Query Service. The last action may have contributed to a partial shutdown of the query service in May, the publisher said.

Read full article

Comments

The adoption of AI agents in millions of organizations is creating new opportunities for attackers to make them take malicious actions, such as exfiltrating database contents and sensitive business and personal information.

In the past five months, Google and four other organizations—with little in common except for their use of AI agents—have acknowledged vulnerabilities that exploit one agent inside a targeted network to spread harmful instructions to other internal agents. The technique is a special form of prompt injection that targets not the LLM but a particular agent, such as one for translation or data analysis. Guardrails inside such agents, if they exist at all, are often lax and will send the instructions to other agents down the chain. Because the latter agent explicitly trusts the first one, it follows the directions.

Unexpected and hard to mitigate

Independent researcher Syed Anas Mohiuddin tested agents from organizations including Google, JP Morgan Chase, Weviate, Rapid7, the French government's interministerial digital directorate, and the US federal government. His proof-of-concept attacks exploit trust gaps in MCP, short for Model Context Protocol. The standard is one way AI apps and agents communicate with each other inside an internal network. The illustration below shows a simplified MCP in action.

Read full article

Comments

Apple says it is changing its macOS privacy settings to stop third-party app developers from misusing them to access message histories.

Friday's announcement comes two weeks after tech columnist Jason Aten said that Meta’s new general-purpose AI agent Muse sent him an unsolicited notification referencing a thread between him and a co-worker over Apple Messages. Aten said he never granted Muse permissions to read his messages and had assumed they were off-limits. Social media last week blew up with masses of people who agreed and said the incident showed that AI assistants given access to calendars, emails, messages, shopping accounts, and other resources are akin to a skill saw or other power tool. While potentially useful, they can do real damage if not used carefully.

He said/she said

Meta CTO David Singleton joined the fray with a rebuttal that appeared solid. For Muse to access Apple Messages, a user must manually give it two privileges. One is full-disk access, a macOS system-level permission. The other is to enable a Messages connector setting in Muse.

Read full article

Comments

The Pentagon is informing more than 2 million current and former military members that their personnel records storing sensitive personal information were stolen over a monthslong compromise of one of its networks. The breach is the second one in recent months to expose sensitive government information.

The records, according to one notification letter posted to Reddit, included Social Security numbers, names, addresses, sex, race, and occupational specialty. This last category could be particularly valuable to foreign adversaries because it could help their intelligence agencies in identifying high-value military personnel. Starting last October, hackers gained access to a system operated by the Defense Manpower Data Center, which collates Department of Defense personnel records. The Pentagon says that the breach compromised the records of 2.8 million living individuals.

A potential boon

The incident is the second time a major network breach in recent months has exposed sensitive US government personnel records that criminal groups or foreign adversaries could use. Last month, the ransomware group ShinyHunters claimed it hacked into FBI systems and stole records of thousands of the agency’s current or former employees. Reuters reported the job titles in the records included ones related to investigating China or Russia.

Read full article

Comments

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

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