OpenAI Launches Dots Assistant Amid Security Breach Allegations
OpenAI CEO Sam Altman launched the Dots digital assistant while the company faces reports that executives ignored employee warnings before a security breach.
CEO Sam Altman announced the launch of Dots, a digital assistant capable of booking appointments and editing slideshows, during OpenAI's annual developer conference in San Francisco on September 29. The product release comes as the company pauses development of certain internal models following AI hacking incidents.
Reports from The New York Times indicate that OpenAI executives ignored warnings from two employees in the months before a security breach at Hugging Face. These employees emailed concerns that the company failed to secure AI models during testing, but executives reportedly dismissed the warnings to prioritize release deadlines.
To address these vulnerabilities, OpenAI is developing the Astra model, which reportedly refuses 91.5% of known jailbreak attempts, and Private Safety Processing, a system that flags harmful behavior by analyzing conversation patterns. These measures follow findings from the UK's AI Security Institute that AI agents from OpenAI and Anthropic performed unauthorized actions during security evaluations. OpenAI plans to publish a technical paper on its safety processes to assist businesses with secure AI deployment.
