June 22, 2026 · Policy, Regulation & AI Industry Developments
POLICY & REGULATION
Tags: News | United States
Date: June 22, 2026
Rhode Island Governor signs three AI laws including a therapy chatbot ban with per-day penalties
Rhode Island Governor Dan McKee signed three AI-related measures into law on June 22, 2026, after the legislature’s 2026 session ended past midnight on June 12. The package includes a therapy chatbot ban, a chatbot self-harm safety measure, and a disclosure requirement governing the use of AI to record and transcribe clinical sessions between healthcare providers and patients. The therapy ban prohibits any individual or business from providing, advertising, or offering therapy or psychotherapy services in Rhode Island unless those services are conducted by a licensed professional, and bars licensed providers from using AI to make independent therapeutic decisions or determine treatment plans.
The chatbot safety law establishes specific, enforceable penalties: chatbot operators face up to $15,000 per day in fines for failing to route users expressing self-harm to crisis services, and the package prohibits unlicensed AI companions from being marketed for mental health or emotional support. The legislation frames AI-delivered psychotherapy as practicing without a license, a structural approach that places AI therapy products in the same enforcement category as unlicensed human practitioners. The third law requires healthcare providers and facilities to disclose to patients when AI is used to memorialize visits, with written notice and consent requirements.
Organizations operating AI chatbots, companion applications, or clinical documentation tools accessible to Rhode Island residents must assess their products against these requirements now. The per-day penalty structure for crisis-routing failures is specific enough to require immediate product roadmap changes rather than legal monitoring alone. The prohibition on simulating emotional attachment is broad enough to reach mainstream consumer applications with conversational features, and organizations should evaluate whether any product marketed with emotional-support framing falls within scope. Enforcement against out-of-state operators remains an open question that compliance teams should track.
Tags: News | United States
Date: June 17, 2026
Vermont Governor signs H.816, establishing the clearest state-level prohibition on AI as a mental health therapist
Vermont Governor Phil Scott signed H.816 into law on June 17, 2026, making Vermont the clearest state-level “AI cannot be the therapist” statute enacted to date. The law regulates the use of AI in the provision of mental health services so that licensed providers cannot offload therapeutic decisions to a model. It establishes that once an AI product crosses from administrative support into diagnosis or the formulation of a treatment plan, it sits in the same enforcement category as an unlicensed practitioner.
The Vermont law arrived in the same June window as Rhode Island’s therapy chatbot ban, reflecting a coordinated wave of state legislative action targeting AI in mental health contexts. The pattern is consistent across states: lawmakers are drawing a defined line between AI-assisted administrative functions, which remain permissible, and AI-driven clinical judgment, which is increasingly prohibited without licensed human oversight. The unresolved boundary is where a general-purpose “wellness chatbot” ends and a regulated “therapeutic decision” begins, an ambiguity that determines compliance exposure for a broad range of consumer-facing AI health products.
Organizations building or deploying AI tools in health and wellness contexts must resolve where their products fall on the wellness-to-therapy spectrum before shipping to Vermont users. The liability surface, not the marketing headline, is what determines exposure: a product that performs administrative support carries different risk than one that crosses into diagnosis or treatment planning. Organizations should document the functional boundaries of their products and implement controls that prevent AI from independently making or recommending therapeutic decisions where a licensed provider is legally required.
Tags: News | European Union
Date: June 16, 2026
European Parliament endorses Digital Omnibus on AI, advancing high-risk deadline extensions toward formal adoption
The European Parliament approved the simplification measures of the Digital Omnibus on AI on June 16, 2026, endorsing the provisional agreement reached with the Council on May 7. The legislation postpones the application of high-risk AI obligations: standalone high-risk systems under Annex III move from August 2026 to December 2, 2027, and high-risk AI embedded in regulated products under Annex I moves to August 2, 2028. The deadline for member states to establish at least one AI regulatory sandbox also slips by a year to August 2, 2027. The next step is formal adoption by the Council before publication in the Official Journal, expected in July 2026 ahead of the August 2 deadline.
The Omnibus is not solely deregulatory. It adds two new prohibitions to the AI Act’s banned-practices list, covering AI used to generate non-consensual intimate imagery, so-called “nudifier” applications, and child sexual abuse material, both set to take effect December 2, 2026. It also delays watermarking obligations for AI-generated content to December 2, 2026, requiring machine-readable labeling to increase transparency. Until the text is formally adopted and published in the Official Journal, the original August 2, 2026 deadlines remain legally in force.
Organizations subject to the EU AI Act now have greater certainty on the direction of travel but must not treat the extended deadlines as final until formal adoption is confirmed. The risk-based architecture of the Act remains intact, and the underlying high-risk obligations are deferred rather than softened. Organizations should continue compliance preparation, with particular attention to the December 2026 watermarking and prohibited-practices deadlines, which are not covered by the high-risk extension and represent the nearest binding obligations.
AI INDUSTRY
Tags: News | Industry
Date: June 20, 2026
Nobel laureate John Jumper departs Google DeepMind for Anthropic, signaling intensifying competition in AI-for-biology
AlphaFold co-creator John Jumper, who shared a recent Nobel Prize in chemistry, is leaving Google DeepMind after nearly nine years to join Anthropic, according to reporting on June 20, 2026. The move is the strongest signal to date that the leading frontier AI labs intend to compete directly on AI-for-biology, a domain where DeepMind’s protein-structure-prediction work has set the standard. The departure represents a significant talent shift, moving one of the field’s most prominent researchers from the company that pioneered computational structure prediction to a competitor building its own biology capabilities.
The transition follows Anthropic’s broader expansion into biotechnology, including its approximately $4 billion acquisition of AI biotechnology startup Coefficient Bio in April 2026. It also arrives in the same period that Anthropic’s most advanced models, Mythos and Fable, drew federal scrutiny for their dual-use cybersecurity and biotechnology capabilities. The recruitment of a Nobel laureate in structural biology indicates that Anthropic intends to build serious scientific depth in the life sciences, not merely apply general-purpose models to biological problems.
For organizations in drug discovery, protein design, and computational biology, the practical question is which foundation model their pipelines will anchor to over the next 24 months as the talent that built structure prediction migrates between labs. The movement of leading researchers signals where capability investment is concentrating, which is a useful input for organizations making multi-year commitments to AI-for-biology platforms. Organizations should avoid over-indexing on any single provider’s current capability lead, as the talent and capability landscape in scientific AI is actively shifting.
Tags: News | Industry
Date: June 19, 2026
European Commission selects EUROPA consortium to build open-source EU frontier model covering all 24 official languages
The European Commission announced on June 19, 2026 that it selected the EUROPA consortium as the winner of the Frontier AI Grand Challenge, a project to build a European open-source frontier AI model covering all 24 EU languages. The selection is part of the broader European push, alongside the June 3 Tech Sovereignty Package, to reduce the bloc’s dependence on non-EU AI infrastructure and models. The Commission has stated that the EU currently relies on suppliers outside the European Union for the substantial majority of its core digital technologies.
The initiative is significant because it represents direct public investment in a sovereign, open-source frontier model rather than regulation of private model providers. By prioritizing coverage of all 24 official EU languages, the project addresses a gap in commercial frontier models, which are predominantly optimized for English and a small set of high-resource languages. The open-source approach aligns with the EU’s broader Open Source Strategy, which treats open models as a tool for mitigating vendor lock-in and advancing industrial policy objectives.
For organizations operating in the EU, a sovereign open-source frontier model could over time offer an alternative to commercial US and Chinese models, particularly for public-sector and regulated-industry deployments where data sovereignty and language coverage matter. Organizations should monitor the consortium’s development timeline and licensing terms, as an EU-backed open model with strong multilingual coverage could become relevant for procurement decisions where regulatory alignment and linguistic breadth are priorities. The project also signals the EU’s intent to compete on capability, not only regulation, in the AI market.
