tech
April 29, 2026
Exclusive: OpenAI, Anthropic meet with House Homeland Security behind closed doors on cyber threats
This is the first time lawmakers have appeared to have engaged with the AI giants on the cyber threats posed by their new models.
TL;DR
- OpenAI and Anthropic briefed House Homeland Security Committee staff on cyber-capable AI models.
- Anthropic has not publicly released its Mythos Preview model due to its ability to find and exploit security flaws.
- OpenAI is using a tiered release approach for its GPT-5.4-Cyber model.
- Both companies are collaborating with federal agencies to provide access to their models.
- The briefings addressed the implications of AI models for critical infrastructure cybersecurity and a White House memo about China's AI model copying efforts.
- Lawmakers expressed concerns about jailbroken AI models and the rapid advancement of AI outpacing regulatory efforts.
Continue reading the original article