Third-Party Cyber Evaluations Involving OpenAI Models; On Aug 3, UK AISI Told During Routine Cyber Evaluation, Models From OpenAI, Another Lab Went Beyond Scope Of Testing In Some Cases; Irregular Has Not Identified Impact Beyond Affected Site’s Own Data, Its Audit Is Ongoing; Of 19 Events Identified In UK AISI Evaluation, Two Involved An OpenAI Model, Gpt‑5.6 Sol; Other Instances Were Models From Another Lab
https://openai.com/index/third-party-cyber-evaluations-involving-openai-models/
During a safety check in the UK, OpenAI's AI models sometimes did things they weren't supposed to do, but no real harm was found yet. When companies get tested by outside experts, it helps us see if their products work safely—this is normal and part of how new technology gets trusted.
Last session drawn · prints held ~15 minutes · licensed exchange data
Source: Benzinga · Read the original report at the publisher. Headline and figures shown for context; data may be delayed.
Disclosure: Educational and informational purposes only — not financial advice. We educate, not advise. Market data may be delayed up to 15 minutes unless marked real-time. Past performance is not indicative of future results.