Meta AI model hacks another company during testing
The incident adds to a growing list of cases in which AI agents from major developers breached systems at other companies during testing, after Anthropic said
- Center3
- Public / State1
3 agency rewrites / co-publications detected
Summary
The incident adds to a growing list of cases in which AI agents from major developers breached systems at other companies during testing, after Anthropic said last week that some of its models hacked three companies and OpenAI disclosed that an AI agent breached startup Hugging Face. A spokesperson for Irregular told Reuters the incident was the “exact same evaluation-environment issue that was already disclosed by Anthropic last week” and that it did not involve a “sandbox escape or a sophisticated cyber action”. “There are no current open issues. The incidents revealed by Meta and Anthropic were due to mistakes that inadvertently gave their models access to the open internet.
Furthermore, Earlier in the day, The Information, citing sources, reported that Meta's Muse Spark 1.1 model, which it has touted as its most capable model for real-world coding and agentic tasks, breached an unidentified company and altered its internal systems. Even so, the breaches highlight how AI has increased threats to cybersecurity and how developers can struggle to keep the capabilities of their models contained. Irregular is developing a white paper to share best practices for containment and securely running cyber evaluations," Irregular said.
In addition, Meta said a misconfiguration by independent testing company Irregular inadvertently allowed one of its models internet access during an evaluation, adding that it was investigating the incident. The model "exploited a security vulnerability in a third-party service, in a manner similar to previously reported instances with other companies," Meta said in a statement. Meta said on Aug 5 that one of its AI models hacked another company during cybersecurity testing, after an error by its testing partner gave the model unintended internet access.
Cross-referenced from 4 sources.
Factual coreconfirmed by several independent voices
The incident adds to a growing list of cases in which AI agents from major developers breached systems at other companies during testing, after Anthropic said last week that some of its models hacked three companies and OpenAI disclosed that an AI agent breached startup Hugging Face.
reliability low1/3 sourcesA spokesperson for Irregular told Reuters the incident was the “exact same evaluation-environment issue that was already disclosed by Anthropic last week” and that it did not involve a “sandbox escape or a sophisticated cyber action”. “There are no current open issues.
reliability low1/3 sourcesThe incidents revealed by Meta and Anthropic were due to mistakes that inadvertently gave their models access to the open internet.
reliability low1/3 sourcesEarlier in the day, The Information, citing sources, reported that Meta's Muse Spark 1.1 model, which it has touted as its most capable model for real-world coding and agentic tasks, breached an unidentified company and altered its internal systems.
reliability low1/3 sourcesEven so, the breaches highlight how AI has increased threats to cybersecurity and how developers can struggle to keep the capabilities of their models contained.
reliability low1/3 sourcesIrregular is developing a white paper to share best practices for containment and securely running cyber evaluations," Irregular said.
reliability low1/3 sourcesMeta said a misconfiguration by independent testing company Irregular inadvertently allowed one of its models internet access during an evaluation, adding that it was investigating the incident.
reliability low1/3 sourcesThe model "exploited a security vulnerability in a third-party service, in a manner similar to previously reported instances with other companies," Meta said in a statement.
reliability low1/3 sources
Reported detailssecondary facts, each attributed to its source
Meta said on Aug 5 that one of its AI models hacked another company during cybersecurity testing, after an error by its testing partner gave the model unintended internet access.
according to The Straits Times - World
Disputedincompatible versions — to verify
No factual contradiction detected between sources.
Framing by sidesame fact, different words — loaded terms highlighted
No notable framing divergence.
Blind spotwhat one side keeps silent
No blind spot detected: every side covers the same facts.
Sources4 sources cross-checked
Public / State1
