Version 7

Two approaches to AI governance remain in contest: the AI Futures Project's Plan A proposes a US-China cooperative pause on frontier AI via joint supply controls and data center audits [1][2], while the Trump administration pursues competitive controls — export restrictions, state preemption, and a National AI Legislative Framework [5][23]. Google DeepMind's classified Pentagon deal has driven a sustained employee revolt spanning from a February 2026 worker letter seeking military AI 'red lines' [11] to a July union vote [14], with mainstream coverage now across the New York Times, IBTimes, and The Hill [16][17][11]. TurnTrout (Alex Turner) has moved from his resignation account to a formal enforcement framework for AI government contracts, backed by binding red lines and the 2026 Al Shimari v. CACI legal precedent [19].

The Google DeepMind story shows concretely what happens when voluntary AI safety commitments meet institutional pressure: one lab exits negotiations, another signs, and internal resistance moves from individual resignation to collective action. TurnTrout's legal enforcement framework attempts to resolve this structurally by attaching liability to safety commitments — shifting the question from whether labs will honor pledges to whether they can be held accountable when they don't.

AI governance is contested between two frames that have produced no convergence. The AI Futures Project's Plan A, published July 2026, proposes a US-China cooperative pause on frontier AI development via joint chip supply controls, data center audits, and research sharing [1][2]. Vitalik Buterin has defended Plan A against critics who call it naive, arguing they apply coordination skepticism to a cooperative pause but not to the alternative — an unmanaged AI transition that concentrates power [3]. The Trump administration has not engaged with the cooperative frame: its January 2025 policy centers competitiveness [4], a March 2026 National AI Legislative Framework expands that posture [5], model weight export controls took effect in January 2026 [6], and the White House is reportedly discussing banning frontier-capability open-weight models [7]. At WAIC 2026, Xi Jinping called for a global AI governance framework, international cooperation to prevent loss of control, and launched a new AI alliance [8][9][10] — the first on-the-record Chinese signals compatible with Plan A's premise, though they have not visibly shifted the Trump administration's competitive posture.

The Google DeepMind Pentagon deal has generated sustained opposition from within the company spanning months. The New York Times reported in February 2026 that Google workers sought 'red lines' on military AI explicitly echoing Anthropic's stance, predating what would become a union drive [11]. Separate reporting indicates the deal blindsided DeepMind's own researchers, who learned of it through public channels rather than internal notice [12]. Researcher Alex Turner (TurnTrout) resigned in July 2026 and published a detailed account documenting that CEO Demis Hassabis removed specific prohibitions from Google's 2018 AI principles while publicly claiming 'nothing's changed about our principles,' and that IASEAI cancelled a promised member poll without explanation [13]. Workers subsequently voted to unionize [14][15], and the story drew coverage across IBTimes, The Hill, and the New York Times [16][17][11]. Anthropic presents the contrasting institutional decision: it ended Pentagon negotiations in February 2026 after the Defense Department insisted on blanket 'anything lawful' usage rights, citing irreconcilable conflicts with its redlines on mass surveillance and autonomous weapons [18].

TurnTrout moved from critique to formal proposal on July 18, 2026, publishing an enforcement framework for AI government contracts [19]. It specifies two binding standards: AI may not be used in autonomous targeting systems without identifiable human control over each specific engagement decision, and AI-generated individualized threat assessments drawn from bulk data require particularized subject identification. A seven-person Defense AI Review Body would advise on compliance, with enforcement operating through transparency — override counts in annual reports visible to all covered employees. High-risk applications are restricted to cloud deployments where the company retains monitoring and suspension rights. The framework grounds corporate motivation in the 2026 Fourth Circuit ruling in Al Shimari v. CACI Premier Technology, which affirmed a $42 million verdict against a defense contractor for harms under government direction, establishing that providers face legal liability even when following government instructions [19].

Domestic US regulation involves several distinct conflicts. Nathan Lambert argues Anthropic's campaign against Chinese model distillation is regulatory capture serving commercial interests, and warns the White House is moving toward an executive order banning open-weight models above frontier capability levels [7]. OpenAI advocates 'reverse federalism' — state AI safety laws in California, New York, and Illinois converging into a de facto national standard — which conflicts directly with the Trump administration's state preemption order and its federal framework target [20][21]. An Alignment Forum post argues political will, not technical research, is the main AI safety bottleneck, citing a 3.6:1 researcher-to-advocate ratio and AI industry's seven-to-one advantage over civil society in EU Commission meetings [22].