Public Policy & Governance
Top Line
A US federal district court issued a preliminary injunction blocking the Department of Defense from designating Anthropic a 'supply-chain risk' in retaliation for the company demanding contractual language barring mass surveillance use of its AI models, though a DC Circuit panel has since declined to pause the label during appeal.
An Ohio man became the first person convicted under a new federal AI statute, pleading guilty to producing AI-generated sexually explicit images including digital forgeries of child sexual abuse, establishing enforcement precedent for AI-specific criminal laws.
Elon Musk's xAI filed suit against Colorado seeking to block the state's algorithmic discrimination law set to take effect in June, claiming the law requiring AI systems to protect residents from discrimination in employment, housing, healthcare and other sectors infringes on First Amendment rights.
A senior Pentagon official overseeing AI efforts, Emil Michael, realized up to $24 million in profit selling his stake in Musk's xAI after the Department of Defense entered an agreement with the company, raising federal conflict-of-interest questions about officials benefiting financially from actions taken in their roles.
OpenAI shelved its Stargate UK investment project citing high energy costs and regulation, undermining the UK government's AI growth strategy just months after announcing a £31 billion commitment as part of a UK-US AI deal.
Key Developments
Federal Courts Split on Pentagon's Retaliation Against Anthropic Over Surveillance Safeguards
On March 16, the US District Court for the Northern District of California issued a preliminary injunction blocking the Department of Defense from maintaining its designation of Anthropic as a 'supply-chain risk' Center for Democracy and Technology. The designation was issued in retaliation against the company for insisting on contractual language that would prohibit the use of its AI models for mass surveillance applications. However, a three-judge panel at the DC Circuit has since declined to temporarily block the Trump administration's move to punish the AI startup during appeal proceedings Politico.
The conflicting rulings expose fundamental tensions between executive branch procurement authority and vendor rights to refuse contracts with surveillance-enabling terms. The supply-chain risk designation mechanism, typically reserved for genuine national security threats, appears to have been weaponized to punish a company for imposing ethical constraints on government AI use — a precedent that could chill other vendors from implementing similar safeguards.
First Federal Conviction Under AI-Specific Criminal Statute Sets Enforcement Template
James Strahler II, a 37-year-old Ohio man, pleaded guilty to cyberstalking and producing obscene visual representations using AI-generated sexually explicit images, including digital forgeries of child sexual abuse The Guardian. The Department of Justice identified this as the first conviction under a new federal AI statute specifically addressing synthetic intimate imagery.
The case provides the first judicial interpretation of how AI-specific criminal provisions will be applied, particularly regarding the distinction between real and AI-generated imagery in existing obscenity and child exploitation frameworks. The guilty plea avoids establishing case law on key definitional questions — such as whether AI-generated imagery depicting non-existent individuals constitutes child sexual abuse material under existing statutes or requires the new AI-specific provisions.
Colorado Faces First Amendment Challenge to Algorithmic Discrimination Law
Elon Musk's xAI filed suit against Colorado seeking to block enforcement of the state's new AI law, scheduled to take effect in June, which imposes requirements on AI systems to prevent algorithmic discrimination in education, employment, healthcare, housing, and financial services The Guardian. The company claims the regulatory framework infringes on its First Amendment rights.
Colorado's law represents one of the most comprehensive state-level attempts to regulate AI system outputs for discriminatory effects rather than merely requiring impact assessments or transparency disclosures. The First Amendment framing is notable — xAI is essentially arguing that algorithmic outputs constitute protected speech and that mandating non-discriminatory outcomes is a form of compelled speech. This mirrors arguments previously made against content moderation mandates, but applies them to decision-making systems rather than communications platforms.
Pentagon AI Official's Financial Windfall From xAI Raises Conflict-of-Interest Questions
Emil Michael, a senior Pentagon official overseeing the Department of Defense's AI efforts, realized profits of up to $24 million from selling his stake in Musk's xAI earlier this year, after the Pentagon entered into an agreement with the company The Guardian. Government ethics records released this month show his stake was valued at a maximum of $1 million when he joined the department. Experts stated that federal law bars officials from taking actions in their jobs that benefit their own financial interests.
The ethics records indicate a 24-fold increase in the value of Michael's xAI holdings during his tenure overseeing Pentagon AI policy, followed by liquidation after the department formalized its relationship with the company. The timing raises questions about whether Michael's oversight responsibilities and the Pentagon's xAI agreement constituted prohibited self-dealing, and whether required recusals were observed.
OpenAI Shelves UK Investment Citing Energy Costs and Regulation
OpenAI has put on hold its Stargate UK investment project, which was part of a £31 billion UK-US AI deal announced in September, citing high energy costs and regulatory burdens The Guardian. The decision undermines the UK government's strategy to position AI at the centre of its economic growth plans.
The withdrawal is notable because it reverses a high-profile commitment made just seven months ago with explicit government endorsement. OpenAI's cited reasons — energy costs and regulation — point to structural competitiveness issues rather than project-specific problems. This mirrors similar concerns that have limited European AI infrastructure investment relative to the US and Middle East, where energy is cheaper and regulatory frameworks are lighter or non-existent.
Signals & Trends
European Commission Initiates AI Energy Consumption Measurement Consultation
The European Commission opened a targeted consultation running until May 15 seeking stakeholder input on measuring energy consumption and emissions of general-purpose AI models and systems European Commission. This consultation is part of a broader study on AI energy efficiency. While measurement standards may seem procedural, they are prerequisite for any enforceable energy consumption mandates under the AI Act or other EU environmental legislation. The timing — concurrent with OpenAI citing UK energy costs as a barrier to investment — suggests the EU is preparing to impose energy efficiency requirements that could further differentiate its regulatory approach from competitors. The consultation represents a shift from regulating AI outputs and processes to regulating AI inputs and resource consumption.
Palantir Engineers Granted NHS Email Accounts With Directory Access
Engineers working for Palantir have been given NHS email accounts, granting them access to a directory containing contact details of up to 1.5 million health service staff The Guardian. NHS.net accounts typically provide access to the entire staff directory. This arrangement blurs the line between vendor and institution in ways that raise data governance questions — particularly regarding who counts as an 'insider' with legitimate access versus an external contractor. The practice suggests that public sector AI adoption is outpacing institutional capacity to maintain clear boundaries between government personnel and private sector AI vendors, with potential implications for data protection compliance and institutional independence.
Civil Society Mobilizes to Preserve AI Act Integrity During Omnibus Amendments
The Center for Democracy and Technology Europe joined 32 civil society organisations in a public letter raising concerns about proposed changes to Annex I of the AI Act during ongoing trilogue negotiations on the AI omnibus Center for Democracy and Technology. Annex I lists product safety legislation relevant to high-risk AI system classification. The mobilization suggests that amendments being negotiated through the omnibus process may narrow the Act's scope or create carve-outs that undermine its integrity. The use of omnibus legislation to modify recently enacted framework laws is procedurally unusual and limits scrutiny — civil society appears concerned that the AI Act is being quietly amended before it has even been fully implemented.
Explore Other Categories
Read detailed analysis in other strategic domains