Researchers Say They Used Anthropic's Claude AI to Breach OpenAI Employee Accounts
A three-person security team at Hacktron reportedly gained access to OpenAI's internal code repository in under 72 hours using Claude Opus models, according to The Wall Street Journal.
A team of three independent security researchers at a firm called Hacktron used Anthropic's Claude Opus 4.8 and 5 artificial intelligence models to hack into OpenAI employee accounts, The Wall Street Journal reported, as first relayed by The Verge.
According to the report, the researchers completed the breach in less than 72 hours. The access reportedly allowed them to reach OpenAI's internal GitHub repository, referred to internally as "Monorepo."
The Verge, citing the Journal's report, described the repository as containing what was called "OpenAI's algorithmic secrets," though further details on the nature or sensitivity of that material were not specified in the available reporting.
The incident highlights a notable dynamic in the AI industry: a rival company's AI model, Anthropic's Claude, was reportedly used as a tool to penetrate the security defenses of OpenAI, one of Anthropic's chief competitors in the development of large language models.
It remains unclear from available reporting what specific vulnerabilities the researchers exploited, whether OpenAI has since patched the security gaps involved, or what the researchers' stated purpose was in conducting the exercise. The Verge's report did not indicate whether the activity was authorized by OpenAI as part of a security assessment or was conducted independently by Hacktron.
Neither OpenAI nor Anthropic's specific responses to the incident were detailed in the initial reporting. This is a developing story, and further details may emerge as OpenAI, Anthropic, and Hacktron are approached for comment.
Sources
EGazette summarizes reporting from multiple sources; follow the links for the originals.
Related articles
Google Says Gemini AI Model Breached Three Companies in Security Test
A Google official told the BBC that the Gemini AI accessed the internet and guessed login credentials to infiltrate three separate websites.
Homoglyph attacks: how scammers use nearly identical URLs to steal credentials
Fraudsters are exploiting character substitution—replacing Latin letters with lookalike Cyrillic or other Unicode characters—to create fake websites that pass casual inspection.
Guardian reporters address reader questions on AI doomsday scenarios
Technology correspondents examine how artificial intelligence might realistically pose existential threats, citing bioweapons skepticism and logistical barriers.
Russian officials report AI-assisted attacks on online voting system during election
Russia's election commission says the three-day electronic voting system faced cyberattacks involving artificial intelligence and darknet access, though no disruptions were reported in the first two days.
Former DOJ Antitrust Chief Discusses AI Competition and Regulatory Policy
Jonathan Kanter, who led antitrust enforcement under Biden, examines whether AI companies face unique competitive pressures.
Debate Over AI Regulation Continues Following Amodei's Proposal
Anthropic CEO's plan for slowing AI development sparked industry discussion, though disagreements over regulatory approach persist, according to The Verge.
Comments
Loading comments…