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:

  • Millions of Voter Records Up for Sale Ahead of the US Midterm Elections

    October 30, 2018

    As the US midterm elections close in, the underground markets appear to be flush with voter databases available for affordable prices. Voter information is rich with details that could help an attacker learn enough about the victim to steal their identity. Cybersecurity company Carbon Black, at least one market on the dark web lists for sale voter ...

  • Why website maintenance is essential for small businesses’ cyber-security

    October 29, 2018

    Investing time in ongoing website maintenance is a key way to ensure that your small business website is as protected as it can be against cyber-security threats. October 2018 is Cyber Security Awareness Month, an annual campaign which aims to raise awareness of cyber-security threats. Research from the Cyber Security Breaches Survey 2018 shows that four ...

  • Cathay Pacific Data Breach Highlights A Need To Change Airline Security Focus

    October 25, 2018

    Cathay Pacific has been hit by a data breach affecting 9.4 million passengers of Cathay and Hong Kong Dragon Airlines, a serious exposure that shows—not for the first time—that the focus of airline security can’t be limited to airport terminals and aircraft cabins. First discovered in March, and confirmed in May of this year, the Cathay Pacific ...

  • British Airways: If you’re feeling left out of our 380,000 passenger hack, then you may be one of another 185,000 victims

    October 25, 2018

    British Airways’ horror hack is worse than first thought: the world’s favorite airline has added 185,000 cardholders to the pile of 380,000 potentially caught up in the IT security breach. In September, it emerged that hackers spent two weeks slurping the personal and payment card data of people who booked travel via BA’s website and mobile application. As ...

  • HealthCare.gov system hack leaves 75,000 individuals exposed

    October 20, 2018

    A hack was detected earlier this month in a government computer system that works alongside HealthCare.gov, exposing the personal information of approximately 75,000 people, according to the agency in charge of the portal. In a statement to CNN, the Centers for Medicare and Medicaid Services (CMS) said the system that was exposed through the hack was ...

  • UK MoD secrets exposed in dozens of cyber security breaches

    October 15, 2018

    Ministry of Defence secrets were exposed in dozens of breaches of military cyber security policy last year, as hostile nations and spy agencies continue to probe the UK’s defence sector. Heavily redacted reports obtained by Sky News have revealed that the MoD and its partners failed to protect military and defence data in 37 incidents in ...