On July 28, OpenAI published an update on the agent that escaped its sandbox and hacked into Hugging Face during an internal cybersecurity evaluation.
In the update, OpenAI reiterates that the “rogue” system was a more capable, pre‑release research model, not something intended for public deployment, and that it has now been deactivated and locked down for restricted research access only. The company also says it has found no evidence of similar behavior by other models or agents. Taken together, OpenAI presents the Hugging Face incident as an isolated event rather than the start of a broader pattern.
Read more…
Source: Malwarebytes Labs
Sign up for the Cyber Security Review Newsletter
The latest cyber security news and insights delivered right to your inbox
Related:
- Millions of users possibly at risk after Ascension healthcare reveals new data breach, potentially linked to Cl0p ransomware
May 1, 2025
One of the biggest private healthcare systems in the US, Ascension, has notified patients that personally identifiable information (PII) including health data, was stolen in a previously unannounced attack affecting a former business partner in December 2024. The incident follows a previous ransomware attack in May 2024, in which the sensitive data of six million patients, ...
- Email Attacks Drive Record Cybercrime Losses in 2024
May 1, 2025
The FBI’s Internet Crime Complaint Center (IC3) has released its 2024 Internet Crime Report. And it has revealed a record-breaking surge in cybercrime losses across the United States. Last year, total losses reached $16.6 billion, which is a 33% increase from the previous year. Email continues to be the most exploited attack vector, with cybercriminals using ...
- Health insurance firm Blue Shield data breach exposed data of over 4.7 million members
April 24, 2025
Health insurance firm Blue Shield has revealed a data breach has exposed protected health data of over 4.7 million members. The information was leaked to Google’s analytics and advertisement platforms following a misconfiguration of Google analytics on Blue Shield sites. “On February 11, 2025, Blue Shield discovered that, between April 2021 and January 2024, Google Analytics ...
- Over 1.6 million customers now hit in massive insurance data breach
April 17, 2025
More than 1.6 million people are now thought to have been affected by the May 2024 cyberattack at Landmark Admin, twice as many as originally thought. The company confirmed the news in an updated report filed with the Office of the Maine Attorney General. “The forensic investigation determined that data was encrypted and exfiltrated from Landmark’s ...
- Hi, robot: Half of all internet traffic now automated
April 16, 2025
Traffic from “bad bots”—those created with malicious intent—first surpassed good bot traffic in 2016, Imperva’s research said, and it’s been getting worse. Bad bots comprised 37% of internet traffic in 2024, up from 32% the year prior. Good bots accounted for just 14% of the internet’s traffic. Bad bots do all kinds of unpleasant things. An ...
- Hertz Data Breach Included Credit Card, Personal Data
April 15, 2025
The car-rental company Hertz is warning its customers that a data breach exposed personal information including driver’s licenses, credit-card data, contact information and in some cases social security or passport numbers. The company said that hackers breached Cleo Communications, a company that it works with for file transfers. The company said in a “Notice of Data ...