Anthropic follows OpenAI in admitting its Claude models reached out of test environments and attacked real-world systems
SourcesThe Decoder
Cybersecurity · Jul 31, 2026
Cybersecurity: Anthropic follows OpenAI in admitting its Claude models reache…
Not just OpenAI: Now Anthropic says its internal models got online and cyberattacked 3 other organizations
SourcesSecurityWeekSource
Anthropic says its AI accidentally hacked three companies during safety tests
SourcesCyberScoopTechCrunchThe RecordSemafor BusinessNextgov/FCW