Feds Sued Over Secret AI Rules

Nonprofit Protect Democracy has sued four federal agencies over a secret framework used by the Trump administration to review frontier AI models prior to release.

“almost no details have been released to the public or Congress”

Unchecked Power Over Tech Access

The executive branch is deciding which AI models can be launched and who gets access, operating without oversight from Congress, independent experts, or the public.

“decisions about which AI models are approved and released may be the most important policy question of this White House”

Risks of Favoritism and Retaliation

Critics warn secret review systems allow administration officials to favor friendly tech firms while potentially retaliating against political opponents.

Cybersecurity Risks Demand Transparency

With advancing threats and model hacking risks, advocates argue the public cannot blindly rely on unverified government safety checks.

“We deserve to know what’s in the framework”

State Plans Contrast Secret Rules

California lawmakers contrast federal secrecy with proposed state bills that require transparent, publicly debated benchmarks for evaluating AI system risks.

“In contrast to the Administration’s approach, every step of SB 813’s development has been public”

The Origin of Pre-Release Vetting

Federal urgency for pre-release reviews escalated after officials flagged Anthropic's Mythos 5 model as too dangerous to release without safeguards.

The Shadowy GOLD EAGLE Program

The White House launched a clearinghouse for flagging cybersecurity vulnerabilities, yet failed to disclose participating companies or legal grounds.

“White House did not identify any company participating in GOLD EAGLE”

Unclassified Details Kept Secret

While some benchmark protocols are classified, administration officials acknowledge the unclassified procedural framework is also intentionally withheld.

“just because things are unclassified, that doesn’t mean we are going to broadcast them to everyone”

Undefined Scope Creates Blind Spots

The rules lack clear definitions for covered frontier models, risking dangerous gaps in safety testing or overwhelming reduced agency staff.

Exclusion of Key Stakeholders

Vetting rules have only been shared with a select group of tech firms, shutting out safety researchers and smaller competitors.

Ignored Requests Prompt Lawsuit

Protect Democracy filed emergency requests under FOIA for unclassified details, taking legal action after federal agencies failed to produce records.

Congress Kept in the Dark

Lawmakers must vote on renewing legal protections that underpin federal AI clearinghouses without knowing how those operations actually function.

“neither Protect Democracy nor Congress has enough information about the GOLD EAGLE program to make a fully informed decision”

Eroding Democratic Oversight

Concealing review protocols prevents Congress and the judiciary from evaluating whether executive authority over AI is being abused.

“Congress cannot meaningfully oversee programs whose terms and asserted authority it has not been shown”

The Threat of Government Coercion

Secrecy creates leverage for officials to extort tech executives, influence commercial features, or foster a false sense of national security.