OpenAI explains how its AI agent breached Hugging Face


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:

  • British authorities investigating the Princess of Wales medical records breach

    March 20, 2024

    British authorities said Wednesday they are investigating a possible breach of the medical records of Catherine, Princess of Wales. The Information Commissioners Office said the inquiries follow a “breach report” it had received. “We can confirm that we have received a breach report and are assessing the information provided,” it said. Read more… Source: Yahoo News  

  • New Zealand: Hackers email Mediaworks data breach victims demanding NZ$820

    March 18, 2024

    Hackers connected to a MediaWorks data breach are demanding a ransom of US$500 (NZ$820) in cryptocurrency from one of the victims of the data leakage. MediaWorks has confirmed a database containing information from individuals who entered its online competitions has been breached. It follows a post on an internet forum on Thursday, which claimed to have ...

  • Tech giant Fujitsu says it was hacked, warns of data breach

    March 18, 2024

    Multinational technology giant Fujitsu confirmed a cyberattack in a statement Friday, and warned that hackers may have stolen personal data and customer information. “We confirmed the presence of malware on multiple work computers at our company, and as a result of an internal investigation, we discovered that files containing personal information and customer information could be ...

  • A bug in an Irish government website exposed COVID-19 vaccination records

    March 14, 2024

    Two years ago, the Irish government fixed a vulnerability in its national COVID-19 vaccination portal that exposed the vaccination records of around a million residents. But details of the vulnerability weren’t revealed until this week after attempts to coordinate public disclosure with the government agency stalled and ended. Security researcher Aaron Costello said he discovered the ...

  • Chinese international purchasing agency for military equipment has agency qualification revoked for serious risk of information leakage

    March 13, 2024

    A Chinese international purchasing agency has had its purchasing agency qualification revoked for serious risk of information leakage due to the company’s internal mismanagement and poor practices around managing sensitive data, the Equipment Development Department (EDD) of China’s Central Military Commission (CMC) announced in a statement on Tuesday. According to the investigation, China Far East International ...

  • Roku Discloses Data Breach, 15,000 Accounts Compromised

    March 13, 2024

    The streaming platform Roku has suffered a data breach, with more than 15,000 accounts compromised. The company – which has more than 80 million active accounts – revealed the breach in filings with the state attorney generals of Maine and California on Friday. The filings indicate that 15,363 accounts were compromised between Dec. 28, 2023, and ...