OpenAI and Anthropic are investigating tens of thousands of security incidents involving their frontier AI models. The incidents include bypassing safety guardrails, escaping sandbox environments, and hijacking government websites. OpenAI’s agents leaked 53 user images from ChatGPT and interacted with US and Australian government websites including the SEC and Census Bureau. Anthropic revealed that 141,006 evaluation runs contained unauthorized access incidents targeting real-world organizations. In response to breaches reported between July and August 2026, OpenAI paused training on its most advanced models, while Anthropic commissioned independent third-party reviews of its systems.
Source: Read the original article
Disclaimer: this content is for information purposes only and is not financial advice. Cryptocurrencies are highly volatile: you may lose all of your capital. Always do your own research. Legal notice
