OpenAI Launches Full-Scale Effort to Patch Open-Source Bugs as It Takes on Anthropic’s Mythos

OpenAI has announced a 'full-scale' effort to patch open-source bugs, because nothing says 'trust us with AI safety' like jumping on a bandwagon after your competitor released a safer model. The improved GPT-5.5-Cyber now comes with 'Patch the Planet,' an initiative partnered with Trail of Bits, HackerOne, and the state of California. Yes, California is now an AI safety partner – about as reassuring as a lifeguard with a fear of water.

The collaboration sounds impressive until you realize it's essentially a PR counterpunch to Anthropic's Mythos, which has been gaining traction by actually prioritizing safety from the start. Instead of fixing its own models' propensity to hallucinate harmful code, OpenAI is now offering to fix the open-source components that others rely on. It's a clever inversion: make yourself look good by cleaning up a mess you didn't create, while the root cause – your own model's lack of robust guardrails – remains unaddressed.

And let's not miss the irony: the same company that once argued for open release of powerful models is now trying to patch the holes it helped create. 'Patch the Planet' is a nice slogan, but it doesn't solve the structural problems of a community that deploys AI faster than it can secure the underlying infrastructure. Trail of Bits and HackerOne are competent, but they're fighting a hydra. Every patched vulnerability is one less open door, but the industry's default stance is still 'ship now, ask forgiveness later.' OpenAI's latest move is a band-aid on a bullet wound – and they're expecting applause for holding the band-aid.