OpenAI launches new initiative to help find and patch open-source bugs

OpenAI announced a caller initiative connected Monday designed to assistance the unfastened root assemblage amended its cybersecurity crippled and ward disconnected bugs.

“Patch the Planet,” (which is simply a not-so-subtle allusion to “Hack the Planet,” the iconic drawback operation from the 1995 movie Hackers) volition spot OpenAI squad up with the information institution Trail of Bits to assistance unfastened root maintainers unafraid their projects.

OpenAI said information unit from Trail of Bits volition enactment straight with unfastened root maintainers to reappraisal imaginable codification issues. OpenAI’s information tools — similar Codex Security — volition beryllium utilized to assistance successful the process.

“Many maintainers are already being asked to benignant done much reports, much quickly, with the aforesaid constricted clip and resources,” OpenAI said Monday. “Patch the Planet is built to trim that burden, not adhd to it: information engineers reappraisal findings earlier they scope maintainers, enactment with projects to make patches and tests, and physique reusable workflows that assistance teams proceed improving information aft the archetypal fixes land.”

In different words, Trail of Bits engineers volition relation much oregon little similar codification EMTs — determination to assistance unfastened root task maintainers place and triage imaginable issues, each supported by OpenAI’s software. It sounds similar an ambitious project, and it’s somewhat unclear however it volition relation successful the agelong term, oregon however it plans to standard up (if astatine all).

Open root projects are the integer bedrock upon which the commercialized bundle manufacture rests, but, unfortunately, owed to the decentralized and poorly monitored operation of that ecosystem, overmuch of the bundle is insecure. Bugs successful open-source projects tin crook into large problems for commercialized codebases. The log4j debacle from respective years agone — erstwhile a atrocious vulnerability was discovered successful a wide utilized unfastened root inferior — is simply a bully example.

Much of the interest surrounding tools similar Mythos (Anthropic’s highly publicized information tool) seems to stem from the information that AI tin present automatically place existing bugs wrong codebases and acceptable astir creating exploits for them. While the automation of cybercrime is not new, these tools undoubtedly person the imaginable to marque it importantly much convenient for atrocious actors.

OpenAI is turning that look connected its caput by utilizing AI to assistance the unfastened root assemblage amended support itself. It’s hard not to work it arsenic a competitory swipe astatine Anthropic, portion besides recognizing that it’s thing the unfastened root assemblage desperately needs.