Global Tech Policy Tracker — 2026-05-25

Top Story

Trump White House Abandons AI Safety Review Mandate After Big Tech Pushback

The most consequential tech policy story of the week centers on a draft AI executive order that the Trump administration prepared but ultimately declined to sign. Politico obtained the draft and reported that the White House had been weighing controls on advanced AI models — including a potential requirement that the federal government vet frontier AI models before release — a significant shift from the administration's previously hands-off posture. The deliberations had "multiple people cautioned remain in flux" and represented what observers called a significant departure from the laissez-faire approach championed by venture capitalists like David Sacks and Marc Andreessen.

But by May 21–23, Trump postponed and ultimately did not sign the order after reportedly disliking "certain aspects" of the draft — specifically provisions that would have imposed a safety review framework on new AI models. According to The Guardian, big tech companies successfully lobbied to have mandatory safety reviews stripped from the draft, calling the reversal a "green light for tech's unchecked power." A White House official stated the outcome reaffirmed a "longtime commitment" to balancing "advancing innovation and ensuring security."

The unsigned order matters for multiple reasons. First, it confirms that the US federal government remains without a binding regulatory framework for frontier AI models — leaving state laws and voluntary standards as the primary guardrails. Second, former DHS Secretary Mayorkas separately endorsed "voluntary" policies for advanced AI threats, referencing Anthropic's Claude Mythos as an example of models with unprecedented hacking capabilities. Third, the episode illustrates how aggressively the tech industry is shaping federal AI policy in real time, with enormous implications for the global regulatory landscape.

1 sources

politico.com

politico.com

New Legislation & Regulatory Actions

United States: Illinois Frontier AI Transparency Bill Advances

• What happened: A bill targeting powerful "frontier" AI models advanced through the Illinois legislature, aiming to increase transparency and establish safety protocols for the most capable AI systems. Advocates described the bill as "only the first step" toward comprehensive AI governance in the state.
• Who it affects: AI developers, deployers, and enterprises using or building large-scale AI systems with operations connected to Illinois.
• Status: Advanced in state legislature as of May 22, 2026; not yet signed into law.
• Why it matters: Illinois joins a growing number of US states seeking to fill the federal regulatory vacuum. The bill's frontier-AI focus — rather than use-case-specific rules — signals a shift in how states are approaching AI governance.

1 sources

npr.brightspotcdn.com

npr.brightspotcdn.com

United States: Minnesota & South Carolina Sign Kids Social Media Safety Acts

• What happened: Minnesota Governor Walz signed a kids' social media safety act into law, one day after South Carolina Governor McMaster signed the Stop Harm From Social Media Act, according to the Transparency Coalition AI's May 22 legislative update.
• Who it affects: Social media platforms with users in Minnesota and South Carolina, particularly those with minor users. Platforms will face new obligations around content exposure, parental controls, and safety features.
• Status: Both laws signed and enacted as of the week of May 22, 2026.
• Why it matters: The two simultaneous state signings reflect an accelerating national trend of legislating children's online safety at the state level, putting pressure on federal lawmakers and platform companies alike to establish consistent national standards.

EU: AI Act Omnibus Deal Finalizes — High-Risk Deadlines Pushed to December 2027

• What happened: The EU AI Act omnibus simplification package was finalized this week with official reporting confirming: high-risk AI system compliance deadlines move from August 2026 to December 2027, machinery systems are carved out of scope, and new prohibitions on "nudifier" apps have been added.
• Who it affects: All companies deploying or developing AI systems in the EU, especially those building high-risk applications in employment, credit, education, and law enforcement.
• Status: Provisional agreement reached; formal adoption process underway.
• Why it matters: The December 2027 deadline extension gives affected companies over 16 additional months to achieve compliance — a major relief for mid-size enterprises that had scrambled to prepare for August 2026 deadlines. Critics argue the delay shows Europe caving to Big Tech pressure.

Enforcement & Penalties

• EU AI Act (no formal fines yet) → All companies operating in EU: As of May 2026, no formal AI Act fines have been levied because core enforcement deadlines (like the now-delayed August 2026 high-risk deadline) have not yet passed. However, the penalty structure is firmly in place: up to €35 million or 7% of global annual turnover for prohibited AI practices (Article 99), and up to €15 million or 3% for high-risk system non-compliance. These thresholds exceed GDPR's maximum penalties in severity, putting AI Act violations potentially above GDPR breaches in terms of financial exposure for large corporations.

• US (Federal) → Frontier AI Developers: The failure of the Trump administration to sign the proposed AI executive order means there is currently no federal pre-release vetting requirement for frontier AI models. Former DHS Secretary Mayorkas noted in mid-May that the White House had been actively meeting with tech companies to discuss regulation and was "considering executive action to address the security threat posed by these AI models." That consideration did not result in binding enforcement action this week.

Industry Response

• Big Tech (collectively): According to The Guardian's reporting from May 23, major technology companies successfully lobbied the Trump White House to remove mandatory AI safety review provisions from the draft executive order. The outcome was described by critics as a "green light for tech's unchecked power." No individual company was named in public reports as the lead lobbyist, but the result reflects the sector's significant influence over executive branch AI policymaking.

• Transparency Coalition AI: The nonpartisan group published its May 22 weekly legislative roundup, highlighting the simultaneous passage of kids' social media safety laws in Minnesota and South Carolina as a notable acceleration in state-level child protection legislation. The coalition described the Illinois frontier AI bill as "only the first step," signaling continued advocacy for stronger measures.

• EU Industry (implicit): The finalized EU AI Act omnibus — which delays high-risk compliance deadlines and narrows scope by carving out machinery — reflects sustained lobbying by European and US technology companies since the original Act passed. Reuters previously characterized the deal as "Europe caving in to Big Tech." The removal of nudifier apps from acceptable use under EU law, however, shows regulators did push back in one area despite making concessions elsewhere.

Region Scorecard

Analysis: What This Means

• For AI developers building frontier models: The failure of the Trump executive order to materialize means there is still no federal pre-release vetting requirement in the US. Companies should treat this as a temporary reprieve, not a permanent green light — the policy debate is clearly alive, and a future administration or Congress could act quickly. Start building voluntary safety review documentation now.

• For enterprises deploying AI in the EU: The December 2027 deadline extension for high-risk AI systems buys meaningful time, but the penalty structure (up to 7% of global turnover) is now finalized and legally in place. Use the additional 16 months to get compliance architecture right the first time rather than rushing a patch-work solution.

• For platforms serving minors: The simultaneous enactment of kids' safety laws in Minnesota and South Carolina — following similar actions in other states — signals that child-focused tech regulation is becoming a standard state-level policy priority in the US. Platforms should anticipate that a majority of US states will have some form of kids' social media law by 2027 and build compliance frameworks for the most stringent provisions now.

• For startups and SMEs: The continuing federal vacuum in the US, combined with the EU's delayed enforcement timeline, creates a window for innovation — but also regulatory uncertainty. Companies operating transnationally should monitor both the US state patchwork and EU enforcement developments closely, as the compliance burden could spike suddenly once deadlines arrive.

What to Watch Next Week

1. Illinois Frontier AI Bill: Whether the Illinois legislature completes passage of the frontier AI transparency bill and sends it to the governor, which would make it one of the first state-level laws directly targeting the most powerful AI models.

2. US Executive Branch AI Policy: With the draft executive order abandoned, watch for any new White House signals — speeches, regulatory guidance, or agency directives — about how the administration intends to address frontier AI security concerns, particularly around models with offensive cyber capabilities.

3. EU AI Act Formal Adoption Timeline: Monitor the European Parliament and Council for the next procedural steps to formally adopt the omnibus amendments, which will set the official clock on the December 2027 compliance deadline.