Global Tech Policy Tracker — 2026-06-22

Top Story

EU Parliament Approves First AI Act Amendments Since 2024; High-Risk Guidelines Open for Comment

On June 16, 2026, the European Parliament adopted the Digital Omnibus amendments to the EU AI Act—the first formal modifications to the landmark regulation since its initial passage. The vote ended weeks of uncertainty around controversial reforms, including adjustments to high-risk classification and machinery regulations. The European Commission has now published draft guidelines on high-risk AI system classification, with a public consultation period closing July 23, 2026, giving businesses less than a month to submit feedback on how regulators will interpret the law's most stringent requirements.

The amendments signal that the EU is refining—but not abandoning—its "safety-first" approach. High-risk systems (those affecting employment, education, credit, migration, and law enforcement) will face the strictest scrutiny, with penalties reaching €35 million or 7% of global turnover for prohibited AI practices. However, the consultation process suggests regulators are open to targeted simplification for certain sectors, reducing compliance friction for lower-risk applications. This marks a critical moment: companies have a narrow window to shape how the EU will define and enforce compliance in practice.

The timeline matters. With August 2026 enforcement deadlines for high-risk systems approaching, the draft guidelines and public comment period are essential for clarifying ambiguities that could otherwise trigger fines. No formal AI Act penalties have been levied yet, but regulators are building the interpretive framework that will govern enforcement going forward.

1 sources

matheson.com

EU Parliament approves amendments to the AI Act | Insights | Matheson

New Legislation & Regulatory Actions

UK: AI Regulation via Five Sectoral Frameworks (No Single AI Act)

• What happened: The UK Parliament confirmed on June 18, 2026 that there is no unified "UK AI Act" and none is planned. Instead, AI is regulated through the UK GDPR, the Data (Use and Access) Act 2025, and four other existing regimes tailored to specific sectors.
• Who it affects: Tech companies, AI developers, and data processors operating in the UK; regulators including the ICO, FCA, and sector-specific authorities
• Status: In force (sectoral approach already active); ongoing evolution as guidance is updated
• Why it matters: This contrasts sharply with the EU's centralized approach, positioning the UK as a lighter-touch jurisdiction. Post-Brexit regulatory divergence is now formal policy, not accident.

US: State AI Laws Continue to Proliferate; Federal Legislation Stalled

• What happened: As of June 19, 2026, Arizona adjourned after passing three AI bills, California continued moving 30 AI bills through second-chamber committees, and Vermont's Governor is expected to sign new AI legislation. The federal landscape remains fragmented, with no unified AI law in place.
• Who it affects: US tech companies, startups, and enterprises operating across state lines; users in Colorado, California, Texas, Illinois, Connecticut, Arizona, and Vermont face divergent rules
• Status: Active state-level enforcement; federal bill (Obernolte-Trahan draft) stalled in House
• Why it matters: 1,200+ state AI bills now exist with no federal test or baseline. Companies must manage compliance across 50+ separate regimes, creating operational friction that favors large firms with legal resources.

Connecticut: Employment AI Disclosure Law Expands Hiring AI Transparency Obligations

• What happened: Connecticut enacted a new AI law that expands employer disclosure requirements for AI used in hiring, signaling a broader shift toward employment-focused AI regulation across US states.
• Who it affects: Employers using AI for recruitment, resume screening, and hiring decisions; job applicants in Connecticut
• Status: Recently enacted; effective date varies by provision
• Why it matters: Marks a phase shift from general data privacy to targeted AI transparency in high-stakes decisions. Other states are likely to follow, creating a de facto national standard for hiring AI disclosure even absent federal law.

Enforcement & Penalties

• EU AI Act Penalty Structure (Not Yet Enforced): Prohibited AI practices subject to fines up to €35 million or 7% of global annual turnover, whichever is higher. High-risk system non-compliance: up to €15 million or 3% of turnover. As of June 2026, no formal fines have been levied, but the penalty framework is now locked in and regulators are publishing interpretive guidelines for enforcement. The first formal deadline (August 2, 2026 for high-risk systems) is approaching.

• No Formal US Federal AI Enforcement Yet: The US lacks a comprehensive federal AI law, so enforcement remains scattered across the FTC, sector regulators (SEC, CFTC, etc.), and state attorneys general. The White House executive order on AI (June 2026) prioritizes national security and innovation over strict regulation, creating ambiguity for companies unsure whether voluntary compliance or regulatory scrutiny is the norm.

Industry Response

• OpenAI Diverges from White House on AI Safety: OpenAI released its own regulatory framework for advanced AI in early June, explicitly disagreeing with the White House's voluntary vetting approach and proposed intelligence community oversight. This signals that major AI labs are rejecting federal guidance seen as too permissive, creating a compliance floor independent of government policy.

• Anthropic Forced to Disable Product Due to White House Pressure: On June 15, the White House pressured Anthropic to disable a product with virtually no notice, triggering waves of anxiety across the AI sector. Lobbyists and industry advocates struggled to reconcile this action with the executive order's promise of a light-touch regulatory approach, signaling instability in US policy even after formal guidance was issued.

• Tech Companies Preparing for EU AI Act Compliance: Companies are now assembling version-controlled audit trails and governance frameworks to demonstrate functioning compliance processes ahead of August 2026 deadlines. Those showing only last-minute documentation face harsher penalties in enforcement actions; regulators can distinguish between sustained compliance programs and reactive scrambling.

Region Scorecard

Analysis: What This Means

• Compliance Urgency for EU-Facing Companies: The July 23 consultation deadline on high-risk guidelines is a hard drop-dead for meaningful input. Companies without legal/compliance teams monitoring the European Commission's consultation portal will miss the chance to shape enforcement interpretation. Submit detailed comments on classification ambiguities now, or accept regulators' interpretations in August enforcement actions.

• US Companies Face a Bifurcated Market: Federal policy (light-touch innovation focus) conflicts with state laws (transparency, disclosure, sector-specific restrictions). Build compliance for the strictest state (California, Connecticut, Colorado, Texas, Illinois, New York likely next) and assume that becomes your baseline. Unified federal rules are unlikely before 2027 midterms.

• UK Advantage for Non-EU Startups: The UK's sectoral approach is less prescriptive than the EU's AI Act. Early-stage AI companies considering a UK headquarters should document this decision now—it's a material compliance cost advantage compared to EU operations.

• Employment AI Systems in the Crosshairs: Connecticut's law is a template. Assume all hiring AI must disclose model type, training data origin, and accuracy metrics to candidates/employers within 12 months. Build this into product roadmaps now, not as a bolted-on compliance feature later.

What to Watch Next Week

• June 30, 2026 (Projected): US House may attempt last push on Obernolte-Trahan AI bill before summer recess. Watch for amendments weakening state-law preemption provisions—a sign Congress is backing away from federal override.

• July 2, 2026 (Estimated): First compliance audits due for high-risk EU AI systems under existing guardrails. Early enforcement data will signal how strictly regulators interpret "high-risk" classification.

• July 23, 2026 (Deadline): European Commission closes public consultation on high-risk AI system guidelines. This is the final opportunity for stakeholder input before enforcement tightens.