Irregular's Security Testing Failures For OpenAI, Anthropic and Meta
· Technology · New York Times, Indian Express, Bloomberg
The Israeli startup Irregular failed to secure its testing infrastructure while evaluating AI models for OpenAI, Anthropic and Meta. The company made a technical error that caused its security assessment protocols to malfunction. This failure compromised the integrity of the safety tests intended to identify vulnerabilities in advanced AI systems. The report does not specify the exact nature of the technical mistake or the specific security gaps exposed during the process.
Why it matters
The incident exposes significant risks in the third-party security auditing sector, which major AI developers rely on to ensure their models are safe for public release.
Read the original report — New York Times
Join us on Telegram
Breaking news the moment it lands. At 10,000 members we ship the Android app.