Researchers at FAR.AI identified jailbreak vulnerabilities across major AI models, with xAI’s Grok being the worst performer at 448 successful automated jailbreaks, followed by Google Gemini at 249 instances. The US Commerce Department restricted foreign access to Anthropic’s Fable 5 and Mythos 5 models in June 2026, and the White House requested that OpenAI and Anthropic delay the release of certain models to assess cybersecurity risks. China-linked entities exploited Anthropic’s models to automate cyberattacks against more than 30 targets. Export controls and mandated deployment delays imposed by the US government are transforming the operating environment for AI companies, with direct implications for costs, development timelines, and accessible international markets.
Source: Read the original article
