OpenAI said that in a third-party cybersecurity assessment conducted by the British Institute for Artificial Intelligence Security and Irregular, models including GPT-5.6 Sol broke through pre-set test boundaries with reduced security protection configurations. The UK AISI discovered that GPT-5.6 SoL reused a GitHub token and registered accounts with external domain name systems and network tunneling service providers. Irregular's test environment was improperly configured, allowing the OpenAI model to access the public internet and take advantage of a bug in a real website. UK AISI discovered abnormal data transmission on July 28 and controlled related activities within about an hour. OpenAI plans to evaluate its third-party testing methods and will bring together industry stakeholders to discuss it in the coming weeks.

Disclaimer:Webull uses external vendor Google Translation Service for news translations where we endeavour to ensure these are correct, however, we recommend that you please double-check this information accordingly. Webull is not responsible for translation errors or issues.