歐盟監管機構依賴OpenAI和Anthropic的自願准入
openai向歐盟監管機構提供了訪問其新模型gpt-5.5 cyber的機會，而委員會仍在努力確保獲得anthropic的mythos模型的類似訪問權。
the european commission confirmed openai offered access to its gpt-5.5 cyber model, welcoming the transparency as a way to oversee deployment and security. commission spokesperson thomas regnier stated that agencies like enisa and the ai office are potential recipients, while discussions with anthropic regarding its mythos model have yet to yield access. this disparity exposes a power imbalance where eu regulators lack legal强制力 to inspect frontier models, relying instead on voluntary corporate compliance. the commission is simultaneously pushing to finalize the cybersecurity act to strengthen its regulatory position.