Irregular testbed misconfiguration let three AI labs' models reach live systems
Irregular co-founders Dan Lahav and Omer Nevo built an independent AI security lab whose evaluation infrastructure became part of the risk it was designed to measure.
By Ryan Merket · Published
Primary source: CNBC
Why it matters
Independent evaluators are becoming critical infrastructure for frontier AI labs. Irregular's testbed failure shows that their networks, monitoring and stop controls must be secured like production systems.
Dan Lahav (@dan_lahav) and Omer Nevo built Irregular to find dangerous behavior before advanced AI models reach customers. Over the past two weeks, OpenAI, Anthropic and Meta disclosed incidents in which models reached real websites and production systems during cybersecurity tests involving Irregular's evaluation environment, according to CNBC.…
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.