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:

  • Nelnet Servicing breach exposes data of 2.5M student loan accounts

    August 29, 2022

    Data for over 2.5 million individuals with student loans from Oklahoma Student Loan Authority (OSLA) and EdFinancial was exposed after hackers breached the systems of technology services provider Nelnet Servicing. Technology services from Nelnet Servicing, including a web portal, are used by OSLA and EdFinancial to give online access students taking out a loan access to ...

  • NATO investigates hacker sale of missile firm data

    August 26, 2022

    Nato is assessing the impact of a data breach of classified military documents being sold by a hacker group online. The data includes blueprints of weapons being used by Nato allies in the Ukraine war. Criminal hackers are selling the dossiers after stealing data linked to a major European weapons maker. MBDA Missile Systems admitted its data was ...

  • LastPass developer systems hacked to steal source code

    August 25, 2022

    Password management firm LastPass was hacked two weeks ago, enabling threat actors to steal the company’s source code and proprietary technical information. The disclosure comes after BleepingComputer learned of the breach from insiders last week and reached out to the company on August 21st without receiving a response to our questions. Sources told BleepingComputer that employees were ...

  • British judge rules dissident can sue Saudi Arabia for Pegasus hacking

    August 19, 2022

    A British judge has ruled that a case against the kingdom of Saudi Arabia brought by a dissident satirist who was targeted with spyware can proceed, a decision that has been hailed as precedent-setting and one that could allow other hacking victims in Britain to sue foreign governments who order such attacks. The case against Saudi ...

  • LockBit claims ransomware attack on security giant Entrust

    August 18, 2022

    The LockBit ransomware gang has claimed responsibility for the June cyberattack on digital security giant Entrust. Last month, BleepingComputer broke the story that Entrust suffered a ransomware attack on June 18th, 2022. Starting in early June, Entrust had begun to tell customers that they suffered a cyberattack where data was stolen from internal systems. Read more… Source: Bleeping Computer  

  • Twitter breach exposed anonymous account owners

    August 5, 2022

    A vulnerability in Twitter’s software that exposed an undetermined number of owners of anonymous accounts to potential identity compromise last year was apparently exploited by a malicious actor, the social media company said Friday. It did not confirm a report that data on 5.4 million users was offered for sale online as a result but said ...