White House Keeps AI Cybersecurity Framework Under Wraps: OpenAI and Anthropic Know the Details, the Public Doesn't

On August 5, 2026, the Trump administration privately shared the details of a new federal cybersecurity framework for AI systems with OpenAI, Anthropic and several other AI labs, keeping the document secret from the general public, Wired reports.

Who got access to the secret document

The framework went to the largest American AI companies — OpenAI, Anthropic and other labs developing advanced models. Independent security researchers, the academic community, nonprofits and the general public still had no access to the text at the time of publication.

Key facts about the situation:

• The document was shared privately on August 5, 2026
• Recipients: OpenAI, Anthropic and other major US AI labs
• No public release has been announced, and there is no official timeline
• Subject: federal-level cybersecurity for AI systems

This approach departs from established American practice for developing security standards. NIST's AI Risk Management Framework (AI RMF), the key federal standard for AI risks currently in force, was published in January 2023 after more than two years of open consultation involving over 200 organizations: technology companies, academic institutions and civil society groups. The 2024 update to the Cybersecurity Framework (CSF) 2.0 also went through a public comment period before becoming a mandatory reference point.

Why researchers criticize the closed process

The secret handoff creates a structural asymmetry: companies with a direct stake in the content of the regulation get early access to the text and a chance to shape its wording, long before the requirements become public. Outside observers — academic researchers, independent security experts, nonprofits, foreign regulators — have no way to verify it independently.

"The public remains in the dark,"

Wired states in its August 5, 2026 report.

In AI cybersecurity, the problem is especially acute. Attacks on AI systems — adversarial examples, deliberate poisoning of training data, jailbreaks of deployed models — have become a recognized threat to critical infrastructure and financial systems. Standards for defending against such attacks have historically benefited from broad external review precisely because developers are not always a disinterested party when it comes to defining minimum security requirements for their own systems.

Closed development also complicates international coordination: regulators in the EU, the UK and Japan, which are actively building their own AI frameworks, can neither study the American document nor check their own standards against it.

According to Wired, the White House has not said when, or whether, the framework will be published openly.

What this means

The Trump administration is choosing to regulate AI through direct, closed coordination with leading labs, bypassing the public consultation that has become the norm for federal technical standards. This continues a course of deregulation and of putting speed ahead of transparency, set after the partial rollback of Biden's executive order on AI in early 2025.

For OpenAI, Anthropic and the other recipients, the document means early access to regulatory requirements and a real chance to help shape the rules. For independent researchers, civil society and foreign partners, it means opacity: neither the framework's contents nor the goals it sets are known. How long the document stays secret will depend largely on pressure from Congress and from civil society organizations interested in transparency in AI regulation.