OpenAI and Anthropic provided a specialized briefing to the U.S. House Committee on Homeland Security regarding their new generation of AI models with cyber attack and defense capabilities and their impact on cybersecurity. This marks the first time the U.S. legislature has conducted a specialized discussion with these two major AI companies regarding the cyber threats posed by new AI models, with a focus on the security risks in vulnerable critical infrastructure sectors. Last Thursday, both companies separately briefed congressional staff in independent, non-public classified meetings. The initiative for these discussions came from Congress, aiming to assess the potential impact of new AI models on the cybersecurity of critical infrastructure. The briefings also referenced allegations in a White House memo accusing China of disassembling and replicating advanced U.S. AI models. OpenAI stated that, in addition to the White House meeting, it conducted multiple similar briefings for various committees in both the House and Senate last week. Anthropic mentioned that last week's meeting was part of its ongoing, regular communication efforts.