Back to Daily Brief

Public Policy & Governance

34 sources analyzed to give you today's brief

Top Line

The UK government has abandoned its preferred position on AI copyright reform after sustained industry opposition, signalling that creative sector lobbying can still block permissive tech-friendly regulatory frameworks even under a pro-growth administration.

The Institute for AI Policy and Strategy has called for targeted US government intervention against Chinese AI companies conducting systematic distillation attacks on American frontier models, marking a shift from voluntary industry countermeasures to potential export control designations.

The FBI has resumed actively purchasing commercial data that can be used to track individuals without warrants, according to Director Kash Patel's first public confirmation — raising questions about compliance with internal policies restricting such acquisitions.

Key Developments

UK Reverses Course on AI Copyright Exception After Creative Industry Pressure

Technology Secretary Liz Kendall confirmed the UK government no longer has a 'preferred option' on copyright reform for AI training, effectively abandoning a previous proposal that would have allowed tech companies to use copyrighted material unless rights holders opted out. The reversal follows sustained opposition from actors, musicians, and writers who argued the opt-out model would undermine their ability to control use of their work, according to The Guardian. The government had framed the original proposal as essential to maintaining the UK's competitiveness in AI development, but faced a coordinated campaign from creative industry unions and trade bodies who threatened legal challenges under existing copyright law.

The policy reversal represents a significant constraint on the Labour government's stated ambition to position Britain as an AI superpower. Unlike the EU's approach in the AI Act — which carved out limited text and data mining exceptions while preserving copyright enforcement — the UK was attempting a more permissive framework that would have shifted the burden to rights holders. The government is now back at square one on what has become one of the most contentious regulatory questions in AI governance: how to balance the compute-intensive data requirements of model training against intellectual property protections that predate generative AI.

Why it matters

Demonstrates that even pro-growth governments face binding political constraints when attempting to override established IP frameworks, suggesting copyright will remain a fragmented jurisdiction-by-jurisdiction battleground rather than converging on a common standard.

What to watch

Whether the government now pursues a licensing-based solution requiring tech companies to negotiate directly with rights holder collectives, or attempts a narrower fair dealing expansion that might survive legal scrutiny.

US Think Tank Proposes Entity List Designations for Chinese Distillation Attacks

The Institute for AI Policy and Strategy has published recommendations for targeted government intervention against Chinese AI companies conducting systematic campaigns to extract capabilities from American frontier models through distillation attacks, citing evidence published by Anthropic, OpenAI, and Google in February 2026, according to IAPS. The memo argues for using Bureau of Industry and Security Entity List designations to restrict adversary AI companies' access to US technology and cloud computing resources, moving beyond voluntary industry countermeasures. Distillation attacks involve querying frontier models millions of times to reverse-engineer their capabilities into smaller, independently deployable models — effectively circumventing export controls on AI chips by stealing the intellectual property encoded in model weights.

This represents an escalation in thinking about AI security controls. Previous measures have focused on chip export restrictions and cloud compute monitoring requirements under the October 2025 Commerce Department interim final rule on advanced computing. Entity List designations would allow BIS to block named Chinese companies from accessing any US-origin technology or services without a license, including API access to closed models and partnerships with US cloud providers. The proposal reflects growing frustration in the national security community that voluntary industry safeguards are insufficient when adversaries are conducting state-backed industrial espionage at scale.

Why it matters

Signals potential shift from compute-focused export controls to entity-specific enforcement targeting firms engaged in systematic IP theft, which could significantly complicate Chinese AI companies' ability to access Western frontier capabilities even without smuggling advanced chips.

What to watch

Whether Commerce Department acts on these recommendations in the next rulemaking cycle, and whether affected Chinese companies attempt to restructure through foreign subsidiaries to maintain access to US cloud services.

FBI Director Confirms Bureau's Active Purchase of Location and Personal Data

FBI Director Kash Patel has publicly confirmed for the first time that the bureau is actively purchasing commercial data that can be used to track individuals, according to Politico. The acknowledgment comes amid ongoing debate over whether law enforcement agencies should be required to obtain warrants before purchasing data that would require a warrant if obtained directly from tech companies or telecom providers. The FBI had previously maintained internal policies restricting such purchases following congressional scrutiny in 2024, but Patel's statement indicates those limitations have been lifted or were never fully implemented.

The legal status of government data purchases remains unresolved. In United States v. Carpenter (2018), the Supreme Court held that accessing historical cell-site location information requires a warrant, but subsequent lower court rulings have split on whether purchasing equivalent data from commercial brokers constitutes a Fourth Amendment search. The FBI's resumed purchasing occurs as the Federal Trade Commission has pursued enforcement actions against data brokers for selling location information to government buyers, and as several states have passed laws restricting commercial data sales. The disconnect between FTC enforcement targeting the sellers and FBI practices as a buyer highlights the absence of a coherent federal framework governing this market.

Why it matters

Exposes a persistent regulatory gap where data that would require judicial authorisation if obtained via subpoena can be purchased commercially without oversight, undermining the warrant requirement and creating pressure for congressional action as civil liberties groups escalate legal challenges.

What to watch

Whether the Wyden-Moran Fourth Amendment Is Not For Sale Act gains traction in this congressional session, and whether any pending circuit court cases produce circuit splits that could push the issue to the Supreme Court.

Signals & Trends

Creative Industry Successfully Weaponises Copyright to Constrain AI Training Regimes

The UK copyright reversal is part of a broader pattern where organised creative industries are using existing intellectual property law as a chokepoint to force AI companies into licensing negotiations. Similar dynamics are playing out in US federal courts where publishers, music labels, and visual artists have filed multiple copyright infringement lawsuits against model developers. What distinguishes this from previous technology-versus-copyright battles is the speed at which rights holders organised — mobilising unions, trade associations, and high-profile talent into coordinated political campaigns within months of generative AI's commercial emergence. Governments attempting permissive training regimes are discovering they cannot override decades of international copyright treaty obligations through domestic regulatory shortcuts, effectively giving rights holder coalitions veto power over national AI strategies.

Export Controls Expanding From Hardware to Intellectual Property and Access

The distillation attack memo reflects a conceptual evolution in technology controls. Traditional export restrictions focused on physical goods — chips, manufacturing equipment, precursor chemicals. The October 2025 advanced computing rules extended this to cloud services and data centre capacity. Now policy advocates are pushing for controls on model access itself, treating API endpoints and training methodologies as controllable items. This creates enforcement challenges orders of magnitude more complex than monitoring chip shipments: how do you prevent a Chinese researcher in Singapore from querying Claude via a VPN? The logical endpoint is entity-specific sanctions rather than technology-neutral rules, which risks fragmenting the global AI research ecosystem and incentivising adversaries to develop fully indigenous capabilities rather than remaining dependent on Western frontier labs.

Explore Other Categories

Read detailed analysis in other strategic domains