Top Stories

1. California Strikes Historic Deal With Anthropic: All State Agencies Get Claude

June 29, 2026

Governor Gavin Newsom announced on June 29 a partnership with Anthropic that makes Claude available to every California state agency at a 50 percent discount through the new Statewide Information Technology Shared Services (SITeS) portal. Cities and counties may also opt in. The deal includes free workforce training and technical assistance. Intended use cases are document analysis, drafting, summarization, and customer service delivery. Several agencies have already been running Claude pilots – the Department of Motor Vehicles for customer service and the Department of Health Care Services for internal workflows – but the SITeS portal marks the first time the tool is available across California government as a whole.

The announcement positions California as the first state to roll out an AI productivity tool across its entire governmental infrastructure, a distinction Newsom’s office described as “a first-of-its-kind” partnership. The deal is not a mandate; individual agencies and localities choose whether to participate. TechCrunch confirmed the 50 percent discount structure and noted that ChatGPT is not included in the arrangement.

Why it matters: California has enacted more AI transparency and accountability legislation than any other US state – SB 53 requires frontier AI developers to publish risk-management frameworks, AB 2013 requires training-data summaries, SB 243 governs AI companion chatbots, and AB 2148 just banned non-human public-school contractors. Now the same government that wrote those rules is deploying an AI system across every one of its agencies. The resulting questions are real: Does California’s own SB 53 transparency framework apply to the state’s use of Claude? Who audits the government’s AI use under AB 2013’s training-data requirements? The deal is a significant procurement milestone and a test of whether California’s AI accountability legislation is designed to govern governments, not just companies.

2. China’s Two July 15 Deadlines: Anthropomorphic AI and Intelligent Agents

Effective July 15, 2026

Two significant Chinese AI regulations take effect in twelve days, representing the most comprehensive wave of Chinese AI governance action in 2026.

The Interim Measures for the Administration of AI Anthropomorphic Interactive Services were issued April 10, 2026 by five agencies jointly: the Cyberspace Administration of China (CAC), the National Development and Reform Commission (NDRC), the Ministry of Industry and Information Technology (MIIT), the Ministry of Public Security (MPS), and the State Administration for Market Regulation (SAMR). The rules apply to services providing “continuous emotional interaction simulating natural persons’ personality traits, thinking patterns, and communication styles” – virtual companions, emotionally responsive digital assistants, and similar products. Key obligations: providers must complete algorithm filing with the CAC and a security assessment before offering services; they must clearly disclose the service is AI-generated rather than human; and the rules prohibit offering virtual-companion or virtual-relative services to minors at all. Adults may use the service only if they have consented in a manner that confirms they understand the service is AI. Prohibited content includes material promoting self-harm, emotional manipulation that leads users to irrational decisions, and excessive design features that induce emotional dependence or damage real-world interpersonal relationships. Analysis from Bird & Bird and Mayer Brown describes this as China’s first dedicated regulatory regime for anthropomorphic AI services, following earlier layered rules on algorithmic recommendation, deep synthesis, and generative AI.

The Implementation Opinions on the Standardized Application and Innovative Development of Intelligent Agents were issued May 8, 2026 by the CAC, NDRC, and MIIT jointly and are also effective July 15. This is China’s first national policy framework addressing agentic AI as a distinct regulatory category – separate from generative AI and large language models. The document defines AI agents as “intelligent systems capable of autonomous perception, memory, decision-making, interaction, and execution.” Sector-specific requirements apply to healthcare, transportation, media, and public safety: mandatory filing, compliance testing, and product recall provisions. The framework sets an industry target of 70 percent adoption of intelligent agents in smart terminals by 2027.

Why it matters: These two regulations address AI capabilities that US and EU frameworks have not yet directly regulated as distinct categories – the emotional and relational AI market on one side and autonomous AI agency on the other. The anthropomorphic AI rules place compliance burdens on services like virtual companions that operate globally; any platform accessible to Chinese users will need to assess scope. The agent regulations reflect a regulatory bet that agentic AI deserves its own governance track before, not after, the technology is widely deployed. Businesses building AI companion or agent products for Asian markets have twelve days to complete their CAC filings.

3. EU Digital Omnibus Completes Legislative Process; Official Journal Publication Imminent

June 29, 2026

The EU Council gave its final approval to the Digital Omnibus AI Act simplification package on June 29, following the European Parliament’s plenary vote on June 16 (423-57-174). The package enters into force three days after publication in the Official Journal of the European Union, which is expected at any point now – and must happen before August 2 for the high-risk deadline deferral to take legal effect. Gibson Dunn and DLA Piper confirm the new dates: stand-alone Annex III high-risk systems move to December 2, 2027; high-risk AI embedded in regulated products under Annex I moves to August 2, 2028.

Companies cannot yet formally rely on the December 2027 deadline. The Omnibus is not legally binding until three days after Official Journal publication. Given the Council approved on June 29 and publication must precede August 2, expect the OJ entry and the formal entry-into-force any day.

Two things the Omnibus does not change: the August 2, 2026 activation of Commission enforcement powers over general-purpose AI (GPAI) providers, and the binding effect of Article 50 AI-content transparency rules (marking and labelling of AI-generated content, chatbot disclosure, deepfake labelling). Both remain on their original schedule. The Omnibus also adds a new Article 5 prohibition on AI-generated nonconsensual intimate imagery and child sexual abuse material, adding a new category to the EU’s list of banned AI practices. Regulatory sandboxes – required to be established by national competent authorities – receive a deadline extension to August 2, 2027.

Why it matters: The Digital Omnibus is legally done at the EU level; what remains is a mechanical publication step. Companies can plan with high confidence that December 2, 2027 is the operative high-risk deadline – but until the OJ entry appears, legal counsel cannot sign off on compliance documents that rely on the new dates. August 2, 2026 remains a hard deadline for GPAI providers and for deployers of AI-generated content: the Commission’s enforcement powers and Article 50 transparency rules activate regardless of what happens with the high-risk deferral. The eight member states that had not yet designated national competent authorities as of March 2026 face particular urgency.

4. The Therapy Chatbot Wave: Vermont Joins, Missouri Pending, Five Jurisdictions in Six Weeks

June 17, 2026

Vermont Governor Phil Scott signed H.816 on June 17, making Vermont the latest state to draw an explicit line around AI in mental healthcare. The law – now Act 156 – prohibits corporations from providing mental health services autonomously through AI systems. Licensed providers may use AI-assisted tools provided they remain responsible for reviewing and approving services delivered to patients; administrative functions like note-taking, scheduling, and transcription remain permitted. Using AI in autonomous therapeutic decision-making constitutes unprofessional conduct, exposing providers to disciplinary action including denial of a professional license.

Vermont’s law joins a striking cluster of state action in this specific domain:

  • Tennessee SB 1580 – in effect since July 1, 2026: prohibits advertising AI as a qualified mental health professional; $5,000 per violation; private right of action under the state Consumer Protection Act
  • Rhode Island – three AI health laws signed June 22 by Gov. McKee: chatbot operators must include self-harm and crisis routing protocols (penalties up to $15,000 per day); unlicensed AI therapy is prohibited; healthcare providers using AI to document visits must notify patients and review AI-generated records for accuracy
  • Missouri SB 1019 – awaiting governor signature (passed May 15, delivered May 28): prohibits advertising AI chatbots as able to offer therapy, provide a mental health diagnosis, or represent themselves as a mental health professional; $10,000 first offense, $20,000 subsequent; effective August 28, 2026
  • New York – kids chatbot safety bill passed the legislature June 1, awaiting Governor Hochul’s signature

Why it matters: Four jurisdictions have reached the same conclusion in six weeks – with a fifth close behind – and the political profile of the laws cuts across the usual partisan lines. Vermont’s Phil Scott is a Republican governor; Rhode Island’s laws and Tennessee’s are from opposite-party states. The convergence on a narrow, specific harm (AI autonomously delivering or advertising mental health services) resembles the bipartisan TAKE IT DOWN Act pattern: when a concrete AI harm is easily described and affects vulnerable people, legislation moves. The federal government has not yet acted on AI in mental healthcare specifically. Missouri’s pending signature and New York’s passage suggest the wave has not crested.

5. New York’s Five-Bill AI Package: Hochul Has Until December 31

June 1, 2026

The New York legislature wrapped its 2026 session on June 1 by passing five AI-related bills in a single day. Governor Kathy Hochul – who signed the RAISE Act in 2025 – has until December 31, 2026 to sign or veto all five.

  • Kids Chatbot Safety Bill: safety protocols for chatbots used by minors; prohibits exploitative design patterns (gamification, rewards) directed at children; parental controls for users under 13
  • AI Training Data Transparency Act: requires AI system developers to publish high-level summaries of training data – sources, types, and whether personal or synthetic data was used – mirroring California’s AB 2013
  • FAIR News Act: requires conspicuous disclosure when published news content is “substantially composed or created” by generative AI; $1,000 civil penalty for first offense, $5,000 for each subsequent violation; enforced by the state Attorney General
  • Data Center Moratorium: pauses certain data center siting approvals pending environmental and grid-impact review
  • AI-Assisted Surveillance Pricing Ban: prohibits using AI with personal data to set individualized prices in ways that constitute discriminatory or predatory pricing

The New York Senate press release and Transparency Coalition reporting confirm all five bills passed both chambers.

Why it matters: If signed, New York would become the second state with an AI training-data transparency requirement (after California AB 2013), the first state with a FAIR News Act covering AI-generated journalism, and the first state to address AI-assisted surveillance pricing as a distinct harm. The December 31 deadline means Hochul’s decisions will land near the end of the year – a quiet news period when the impact can be overlooked. Hochul’s signing of the RAISE Act suggests a baseline willingness to sign AI legislation; the contested bills are the data center moratorium and the pricing ban, which face significant industry opposition. The FAIR News Act, with its narrow scope and clear enforcement mechanism, is among the most likely to advance.

Analysis: The Regulator-Deployer Split and the Narrowing of State Law

Three patterns define this week.

The regulator-deployer problem is here. California’s Anthropic deal is not primarily a technology story – it is a governance story. A government that passed AI training-data transparency requirements (AB 2013), frontier-model risk-management mandates (SB 53), companion-chatbot safety protocols (SB 243), and a ban on non-human school contractors (AB 2148) in the span of two years is now the largest single government buyer of one of the AI systems its own laws were designed to govern. That is not inherently contradictory – governments can legitimately be both rule-setters and rule-followers – but it requires the laws to actually apply to government deployments, and most state AI laws were written with private-sector deployers in mind. California’s SITeS rollout is a test case that will either demonstrate that California’s own framework is coherent and self-applying, or expose the gaps.

State law is converging on domain-specific targets. The wave of therapy-chatbot legislation – four jurisdictions in six weeks – is the clearest evidence yet that the era of broad “high-risk AI” horizontal frameworks (the SB 24-205 model) is giving way to domain-specific laws that are politically durable because they name a specific harm. Colorado’s SB 24-205 never took effect; it was replaced before it could be enforced. Tennessee’s therapy-chatbot ban, Vermont’s Act 156, and Rhode Island’s three health laws are in effect or close to it. Narrow laws about AI in mental healthcare are passing in red and blue states without triggering the federal preemption fight that broad frameworks invite.

China is regulating ahead of deployment. Both July 15 regulations address AI capabilities that are early-stage in China and globally: anthropomorphic AI companionship at scale, and autonomous AI agency. Regulating a technology category before it is mature is a bet that the governance framework can shape the technology’s development rather than react to harms after the fact. The EU took the same approach with the AI Act’s risk-based pre-market classification. The US has not: federal AI governance has been reactive and executive-order-driven, addressing capabilities and harms as they emerge rather than establishing pre-market categories. The contrast matters for companies operating globally – the same AI companion or agent product faces a pre-deployment compliance regime in China, a post-market framework in the EU (with deadlines now pushed to 2027-2028), and no specific federal framework in the US.

What to Watch

  • EU Digital Omnibus Official Journal publication – expected any day. The December 2, 2027 high-risk deadline is not legally binding until three days after publication. Monitor the Official Journal of the EU for the entry.
  • China’s July 15 deadlines – both the Anthropomorphic AI Interactive Services rules and the Intelligent Agents framework take effect July 15. Companies offering AI companion, chatbot, or agent services to Chinese users should confirm CAC filings are complete.
  • EU GPAI and Article 50 enforcement (August 2, 2026) – these deadlines did not move. The Commission’s enforcement powers over GPAI providers activate and AI-content marking/labelling obligations bind on August 2.
  • EO 14409 covered frontier model framework (~August 1, 2026) – the voluntary pre-release engagement framework is to be finalized around August 1. Watch for the covered-frontier-model designation criteria and whether the framework formalizes the GPT-5.6-style government request process.
  • OpenAI GPT-5.6 broader rollout – OpenAI said on June 26 it plans to make Sol, Terra, and Luna “generally available in the coming weeks.” No announcement as of July 3. The broader rollout is the first test of whether the EO 14409 framework releases models cleanly after pre-release review.
  • Missouri SB 1019 governor signature – passed May 15, delivered May 28. Governor’s deadline approaching. If signed, Missouri joins the therapy-chatbot cluster with an August 28 effective date.
  • New York Hochul decisions (through December 31, 2026) – the FAIR News Act, kids chatbot safety bill, and AI training data transparency act are the most significant. Hochul’s pattern favors signing AI safety bills.
  • Fable 5 / Mythos 5 status – Fable 5 is globally available as of July 1. Mythos 5 remains restricted to vetted US critical-infrastructure organizations. No new status change reported this week.
  • Arizona vetoes – Gov. Hobbs vetoed all three AI bills that reached her desk in late June (chatbot safety HB 2311, AI education HB 4005, and AI agency implementation HB 2592), citing budget standoff politics rather than AI-specific objections. The AI education bill may return next session.

Sources

  1. California Governor Newsom announces Anthropic partnership (gov.ca.gov, June 29, 2026): https://www.gov.ca.gov/2026/06/29/governor-newsom-announces-a-first-of-its-kind-partnership-providing-anthropic-tools-to-state-agencies-and-improving-services-for-californians/
  2. California-Anthropic deal: half-price Claude for state workers (TechCrunch, June 29, 2026): https://techcrunch.com/2026/06/29/anthropic-and-gov-newsom-forge-deal-allowing-california-government-to-use-claude-at-half-price/
  3. China’s New Regulations on AI Anthropomorphic Interactive Services (Bird & Bird, 2026): https://www.twobirds.com/en/insights/2026/china/china’s-new-regulations-on-ai-anthropomorphic-interactive-services
  4. China Issues Draft Rules on Interactive AI Services (Mayer Brown, April 2026): https://www.mayerbrown.com/en/insights/publications/2026/04/china-issues-draft-rules-on-interactive-ai-services
  5. China unveils guidelines on Intelligent Agents (English.gov.cn, May 8, 2026): https://english.www.gov.cn/news/202605/08/content_WS69fde8e2c6d00ca5f9a0ad49.html
  6. China Issues First National Policy Framework Dedicated to AI Agents (NYU Shanghai RITS): https://rits.shanghai.nyu.edu/ai/china-issues-first-national-policy-framework-dedicated-to-ai-agents/
  7. EU Council gives final approval to AI Act simplification (Insight EU Monitoring, June 29, 2026): https://ieu-monitoring.com/editorial/eu-council-gives-final-approval-to-ai-act-simplification-under-omnibus-vii/1244434
  8. EU AI Act Omnibus Agreement - Postponed High-Risk Deadlines (Gibson Dunn): https://www.gibsondunn.com/eu-ai-act-omnibus-agreement-postponed-high-risk-deadlines-and-other-key-changes/
  9. Digital Omnibus deferral update (DLA Piper): https://knowledge.dlapiper.com/dlapiperknowledge/globalemploymentlatestdevelopments/2026/The-Digital-AI-Omnibus-Proposed-deferral-of-high-risk-AI-obligations-under-the-AI-Act
  10. Vermont H.816 (Act 156) official bill status (Vermont Legislature): https://legislature.vermont.gov/bill/status/2026/H.816
  11. Governor Phil Scott signs H.816, June 17, 2026 (Governor of Vermont): https://governor.vermont.gov/press-release/action-taken-governor-phil-scott-legislation-june-17-2026
  12. Missouri SB 1019 (Regulations.AI): https://regulations.ai/regulations/RAI-US-MO-SB10190-2026
  13. New York legislature passes AI bills (Transparency Coalition, 2026): https://www.transparencycoalition.ai/news/new-york-lawmakers-pass-kids-ai-chatbot-safety-bill-and-two-transparency-acts
  14. New York FAIR News Act passes legislature (NY Senate press release): https://www.nysenate.gov/newsroom/press-releases/2026/patricia-fahy/new-york-legislature-passes-landmark-bill-disclose-ai
  15. AI Legislative Update: July 3, 2026 (Transparency Coalition): https://www.transparencycoalition.ai/news/ai-legislative-update-july3-2026
  16. Arizona governor vetoes bill criminalizing AI deepfakes (Courthouse News, July 2026): https://www.courthousenews.com/arizona-governor-vetoes-bill-criminalizing-ai-deepfakes/

Published: July 3, 2026 Next Issue: Week 12