Government: AI trends to focus on - Courts and states draw harder lines around human control

This week the AI policy conversation moved from principles to practical questions of control and cost. Courts and states are drawing harder lines around human decision-making, while federal leaders debate how—and whether—to impose guardrails. For public sector leaders, the signal is clear: prepare for overlapping rules, document why high-stakes decisions stay with people, and watch infrastructure spending closely.

What changed this week

The US judiciary made its position plain: AI can assist with case management, but judicial decisions remain human. The courts reported progress on AI guidance while explicitly reserving judgment for judges. That line—between administrative efficiency and actual decision authority—is now formal policy in the federal court system.

At the state level, California and Massachusetts accelerated independent oversight. Governor Newsom issued an executive order to speed up independent AI evaluation and advance work on a literal kill switch. Governor Healey called for mandatory independent evaluations and incident reporting in Massachusetts legislation. Colorado lawmakers renewed their push for AI regulation. These are not pilot programs; they are binding frameworks in progress.

The federal picture remains unsettled. White House aides are split on AI policy direction, and political leaders are racing to respond to growing risks. Governor Shapiro urged Congress to establish guardrails. Former FTC Chair Lina Khan argued existing law can address current AI harms, pushing back on calls for sweeping new statutes. Meanwhile, Trump proposed a new AI czar and an AI Force modeled on Space Force—a structural shift that would embed AI oversight in a military-style chain of command.

Infrastructure costs broke into the mainstream debate. Axios hosted a live discussion on how AI data-center growth is intensifying conflict over grid costs. The question of who pays for the energy infrastructure to support AI is no longer theoretical.

On the international front, the EU AI Board focused its ninth meeting on enforcement, frontier incidents and cybersecurity. China pushed back on US warnings about rapid AI development. Governments are treating AI as strategic infrastructure and a source of oversight risk simultaneously.

What it means for you

You are now operating in a multi-layered compliance environment. Federal guidance may stall, but states and courts are not waiting. If your agency uses AI in any decision that affects a person's rights, benefits or legal standing, you need to document why a human made the final call. The judiciary's stance is likely to influence other branches of government and could become a de facto standard in litigation.

Independent evaluation is becoming the norm, not the exception. California and Massachusetts are building models that other states will copy. If you procure AI systems, expect requirements for third-party testing, incident reporting and possibly a kill switch. Start asking vendors how they support these capabilities now.

The infrastructure debate will hit budgets. Data-center energy demands are colliding with grid constraints. If your agency is planning AI deployments that require significant compute, factor energy costs and siting challenges into your timeline and budget projections.

The DOJ's copyright filing on AI took key agencies by surprise, signaling that legal positions on training data and intellectual property are still fluid. If your agency uses generative AI trained on third-party content, legal exposure is not settled. Coordinate with counsel before expanding use.

Finally, the lawsuit alleging an illegal agreement among Anthropic, OpenAI, SpaceXAI and Google to slow AI development adds a new dimension to procurement risk. If your agency relies on models from any of these vendors, monitor the case for potential impacts on product roadmaps and pricing.

What to focus on next week

• Review any AI system your agency uses that influences decisions about individuals. Confirm that a human retains final authority and that the process is documented.
• Check vendor contracts for independent evaluation clauses, incident reporting obligations and kill-switch capabilities. If they are absent, flag the gap for procurement and legal teams.
• Brief your budget office on the emerging grid-cost debate. If AI infrastructure spending is in your pipeline, prepare a short assessment of energy requirements and potential cost exposure.
• Ask your legal team to assess the DOJ copyright filing and the AI slowdown lawsuit for any direct impact on your current or planned AI tools.
• Monitor state-level AI legislation in California, Massachusetts and Colorado. Even if you operate elsewhere, these bills often set templates that spread quickly.

For the full list of stories this week, see all Government AI news.