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:
- Russian State-Sponsored Cyber Actors Target Cleared Defense Contractor Networks to Obtain Sensitive U.S. Defense Information and Technology
February 16, 2022
From at least January 2020, through February 2022, the Federal Bureau of Investigation (FBI), National Security Agency (NSA), and Cybersecurity and Infrastructure Security Agency (CISA) have observed regular targeting of U.S. cleared defense contractors (CDCs) by Russian state-sponsored cyber actors. The actors have targeted both large and small CDCs and subcontractors with varying levels of ...
- Red Cross: State hackers breached our network using Zoho bug
February 16, 2022
The International Committee of the Red Cross (ICRC) said today that the hack disclosed last month against its servers was a targeted attack likely coordinated by a state-backed hacking group. During the incident, the attackers gained access to the personal information (names, locations, and contact information) of over 515,000 people in the “Restoring Family Links” program ...
- Croatian phone carrier data breach impacts 200,000 clients
February 11, 2022
Croatian phone carrier ‘A1 Hrvatska’ has disclosed a data breach exposing the personal information of 10% of its customers, roughly 200,000 people. The announcement does not provide many details other than that they suffered a cybersecurity incident involving the unauthorized access of one of their user databases, which contained sensitive personal information. The type of information that ...
- Fortune 500 service provider says ransomware attack led to leak of more than 500k SSNs
February 5, 2022
Morley Companies, an organization that provides business services to dozens of Fortune 500 companies, said this week it was hit with a ransomware attack last year that led to the leak of sensitive information for more than 500,000 people. In a press release, the company said the ransomware attack began on August 1 and made their ...
- China suspected in hack of journalists at News Corp
February 4, 2022
Digital intruders broke into News Corp email accounts and compromised the data of an unspecified number of journalists, the company disclosed Friday. The media firm’s internet security adviser said the hack was likely aimed at gathering intelligence for Beijing’s benefit. News Corp, which publishes the Wall Street Journal, said the breach was discovered in late January and ...
- What Does an Internal Attack Resulting in a Data Breach Look Like in Today’s Threat Landscape?
February 3, 2022
A common scenario is one in which an attacker gains access to an internal network via a compromised workstation that has been infected with malware, invariably via a social engineering email attack. No enterprise is immune to this type of insider attack. We all, at some point, took the bait and clicked unsolicited links masquerading ...