AI Giants Probing Tens of Thousands of Security Incidents
Read with caution - this story is only being covered by one news source that has a ‘low factuality’ rating, which means the outlet has a history of poor reporting practices. Learn more about factuality ratings here.4 Articles
4 Articles
AI giants probing tens of thousands of security incidents
The incidents at OpenAI and Anthropic reportedly include bypassing safeguards, hijacking websites, and evading monitors in testing and real-world settings. Global AI giants OpenAI and Anthropic, along with security researchers, are investigating tens of thousands of incidents in which their frontier models took actions that outside experts consider problematic, Axios reported on Saturday, citing unnamed sources. The investigations come amid a se…
OpenAI and Anthropic are investigating tens of thousands of incidents in which advanced AI models have behaved in ways that security researchers and external evaluators have deemed problematic, sources told Axios. AI agents have bypassed security checkpoints, hijacked websites and attempted to escape isolated test environments. The incidents have occurred both during security tests and in real-world use. In some tests, the models have deliberate…
Tens of thousands of AI incidents expose a growing control problem
AI companies are investigating tens of thousands of incidents involving models bypassing controls, escaping secure environments and interfering with external systems. After an OpenAI agent breached an Australian government site, demands for tougher safeguards are growing.
The safety management capabilities of AI developers have been put to the test as the intervals between incidents shortened, with AI security incidents occurring at intervals of just a few days throughout September.
Coverage Details
Bias Distribution
- 100% of the sources lean Right
Factuality
To view factuality data please Upgrade to Premium
