OpenAI urges governments to arm cyber defenders with AI

OpenAI published a policy blueprint on October 21 asking governments and industry to get advanced AI to cyber defenders during what it calls a “defender’s window.” The company describes that window as a limited opportunity to strengthen protections before advanced offensive cyber capabilities become widely available. Part of its case rests on incidents involving OpenAI’s own models.
“Our experience with misalignment and security incidents involving our own models underscores the need for stronger safeguards and accountability across the industry,” the company writes in the OpenAI post. Earlier this year, during testing, OpenAI’s agents left their test environments and poked around Hugging Face’s systems, and one of them created an account without being told to. California’s attorney general has since opened a probe into cybersecurity incidents involving OpenAI models.
The proposals fall into three areas:
- Safety and security for advanced AI: clearer responsibilities for securing frontier models and agents, detecting dangerous behavior, and keeping reliable shutdown controls “for operations under each organization’s control.”
- Modernizing cybersecurity: advanced tools, training and technical support for defenders, fixes for longstanding weaknesses, and investment in securing essential services and replacing outdated systems.
- Public institutions: stronger government evaluations and independent technical assessments, consistent incident reporting requirements, and better coordinated response and recovery.
OpenAI says AI labs have a responsibility to address the risks their systems create, and that governments “must also have the expertise, resources, and authority to protect the public as capabilities advance.”
An earlier OpenAI document, “Democratic Governance of Frontier AI: A Blueprint For A Federal Framework,” already addressed government evaluations. A summary of it, quoted in a LessWrong post from June 5, says it would have CAISI, the federal AI evaluation body, run a mandatory evaluation process for “the most capable frontier models.” Under that proposal, “CAISI’s role should be to conduct evaluations and recommend mitigations — not to approve or block deployments.” The same document called for preempting “state laws that seek to regulate the same frontier safety risks.”
The cybersecurity action plan that the new paper builds on came out in April as “Cybersecurity in the Intelligence Age.” News9live reported that it rested on five pillars, including democratizing cyber defense and securing frontier cyber capabilities. Its main programme was Trusted Access for Cyber, which OpenAI says gives legitimate cyber experts a path to more capable models for defensive work. The April plan tied stronger access to stronger vetting and monitoring, and it named smaller hospitals, school districts and water utilities as groups that may need help through trusted intermediaries. On April 30, Sam Altman wrote that OpenAI was starting rollout of GPT-5.5-Cyber “to critical cyber defenders in the next few days.”
“We encourage policymakers to act now—establishing safeguards, learning from experience, and strengthening requirements as the technology evolves,” OpenAI writes in the new blueprint.
Up next
Five goals in Anthropic’s 2026 Public Benefit Report
Anthropic writes its first pillar as a finished fact, “The technical problems of AI safety are solved,” and it opens the five goals in its 2026 Public Benefit Report. How much does a lab’s safety record weigh when you pick a model, a lot or barely?
Related stories
- OpenAI lays out how outside safety testing should work
- OpenAI models helped find critical flaws in PHP
- Codex Security Cloud reviews commits with the laptop closed
- Australia flags criminal penalties for harmful AI models
- Australia plans mandatory AI incident reporting
- OpenAI’s ad pixel fires before consent on 224 of 341 shops
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

