Editorial Team
8 min read
Five states enacted AI mental-health restrictions in 2026, but the laws regulate different conduct: Tennessee targets representations that a system is a qualified professional, while Colorado and Maine restrict how licensed clinicians can use AI in therapy.
States are moving into a federal policy gap around AI and mental health, but the shorthand of an “AI therapy chatbot ban” obscures the central fact: the laws regulate different actors and different conduct. Tennessee's new law is principally a consumer-protection rule about public representations. Colorado and Maine go further into the clinical workflow, setting conditions for a licensed professional's use of AI.
That distinction matters because the underlying behavior is not confined to purpose-built therapy products. A 2025, cross-sectional survey of 1,009 U.S. respondents aged 12 to 21, weighted to 42.8 million people in that age group, found that 19.2% reported ever asking an AI chatbot for mental-health advice. Of those users, 42.8% said they did so at least monthly and 63.3% said they had told no one. The study measured self-reported advice-seeking, not clinical care, safety, or treatment outcomes; it did not identify the specific chatbot a respondent used.
Colorado, Maine and Tennessee enacted their measures in the first half of 2026. A June legislative update grouped those three with 2025 laws in Illinois, Nevada and Utah. A later state-law roundup also identified Rhode Island and Vermont as 2026 enactments.
The five-state count is useful context, not proof that the statutes impose the same test. The clearest contrast is among the three measures with retained primary text:
| State | What the law targets | Operational consequence |
|---|---|---|
| Tennessee | Developing or deploying an AI system that advertises or represents itself to the public as, or able to act as, a qualified mental-health professional. | A violation is an unfair or deceptive act under the state consumer-protection law; the bill summary specifies a $5,000 civil penalty per violation. |
| Colorado | A regulated psychotherapy professional's use of AI, as well as specified consumer-facing representations by any person. | Therapeutic AI communication requires synchronous real-time participation by the client, clinician and system; AI recommendations or treatment plans require clinician review and approval. |
| Maine | Offering therapy or psychotherapy to the public, including through internet-based AI, unless provided by a licensed professional; it also regulates clinicians' supplementary use of AI. | The clinician remains responsible for interactions, outputs and data use; AI may not make independent therapeutic decisions or directly conduct therapeutic communication. |
Rhode Island and Vermont are part of the reported 2026 total, but the retained Rhode Island package contains no article body. The available roundup describes Rhode Island as restricting unlicensed provision or advertising of therapy and barring AI from independent therapeutic decisions or treatment plans, and Vermont as prohibiting independent delivery of mental-health services by AI. Those descriptions should not be read as a substitute for each law's text.
Tennessee's SB 1580, sponsored by Tennessee state Sen. Page Walley, a clinical psychologist, became law on April 1 and took effect July 1. The official bill record says it bars a person from developing or deploying an AI system that advertises or represents to the public that it is, or can act as, a qualified mental-health professional. It passed the Senate 32-0 and the House 94-0.
That is a consequential but relatively narrow lever. It does not, in the language retained with the bill record, create a general prohibition on every chatbot that discusses stress or emotions. Its immediate compliance focus is a product's public claim: marketing copy, interface language, personas and other representations that may say or imply the system is a qualified professional.
The statute routes violations through Tennessee's Consumer Protection Act, bringing the act's remedies—such as injunctions, private actions and damages—alongside the specified $5,000 civil penalty. It leaves the harder line-drawing question for enforcement: when does a general-purpose system's behavior or presentation amount to a representation that it can act as a professional?
Colorado’s enacted HB 26-1195 summary describes limits on clinicians’ use of AI in psychotherapy. Source: Colorado General Assembly.
Colorado's HB 26-1195 became law on June 3 and takes effect August 12. The enacted bill summary covers people lawfully permitted to provide psychotherapy services in the state—regulated professionals—and disciplines them through the relevant licensing board if they violate its limits.
For therapeutic communication, the law requires the client, the regulated professional and the AI system to interact synchronously and in real time. It also prohibits a professional from relying on AI-generated therapeutic recommendations or treatment plans without review and approval. Separately, Colorado makes it a deceptive trade practice to imply that AI output is provided by, endorsed by or equivalent to a regulated professional's services; to represent that the system provides psychotherapy; or to make a misleading therapist-confidentiality-style claim about user data.
But Colorado also limits the reach of its own rule. It preserves disclosed technology that does not diagnose or treat mental-health disorders and is not presented as a substitute for clinical care, including self-help, therapeutic homework, coaching, patient navigation, guided meditation and journaling. It also preserves FDA-regulated technology, certain research, and education or training uses. That is why calling the measure a total ban on AI in mental health would overstate it.
Maine's LD 2082 draws a related line through the definition of therapeutic communication. Its enacted text includes interactions intended to diagnose, treat or address mental, emotional or behavioral-health concerns, including guidance, therapeutic strategies, emotional support, reassurance or empathy in response to distress, treatment planning and behavioral feedback. A person may not offer therapy or psychotherapy to the public through internet-based AI unless the services are provided by a licensed professional; a violation is an unfair trade practice.
For licensed professionals, the Maine law allows administrative work such as scheduling, billing and logistics. It also permits defined supplementary support, including therapy notes, review of anonymized progress data and organization of external resources or referrals. The trade-off is responsibility: the licensee remains fully responsible for interactions, outputs and data use; AI cannot make independent therapeutic decisions, directly interact in therapeutic communication, or generate a therapeutic recommendation or treatment plan without clinician review and approval.
When supplementary AI support is used during a recorded or transcribed session, Maine requires written notice of the tool, its purpose, and how session data will be stored, retained, used for training and deleted when therapy ends, plus consent. Acceptance of general terms, passive interface activity and deceptive conduct do not count as consent. A patient also cannot be refused therapy solely for declining that consent.
Reported use of AI chatbots for mental-health advice in similar 2024 and 2025 U.S. youth surveys; the study notes different wording between surveys. Source: JAMA Pediatrics.
The youth survey gives lawmakers a reason to treat the issue as more than a niche product dispute. It estimated 8.2 million 12-to-21-year-olds had ever used an AI chatbot for mental-health advice, based on the 19.2% weighted estimate. The study also found that use was more common among female respondents than male respondents and among respondents aged 18 to 21 than those aged 12 to 14, after adjustment for the study's listed demographic and geographic factors.
Its most decision-relevant limit is equally important. The researchers did not ask about particular products, did not measure the quality of care, and described the result as a late-2025 snapshot. A high rate of perceived helpfulness—91.7% of users rated advice somewhat or very helpful—cannot establish that advice was clinically sound. The authors specifically noted that perceived helpfulness may reflect a chatbot's agreeable responses rather than the quality of its advice.
This leaves a mismatch between product categories and legal categories. The survey's examples included general-purpose and platform chatbots such as ChatGPT, Google Gemini, My AI, Character.AI and Meta AI. The statutes, by contrast, turn on terms such as therapeutic communication, psychotherapy, professional representation, independent decisions and clinician review. A consumer could use the same broad tool in ways that fall on different sides of those legal lines.
The policy question now is whether states can make their boundaries predictable enough for patients, clinicians and vendors to follow. Tennessee will have to decide what evidence shows a prohibited professional representation. Colorado's August effective date will test whether clinicians can distinguish permissible wellness or administrative uses from regulated psychotherapy activity. Maine's boards must implement rules for a model that puts consent, confidentiality and professional responsibility at the center of supplementary AI use.
The evidence needed next is practical, not rhetorical: regulator guidance, enforcement decisions and information about whether clinicians can continue offering compliant tools while consumer products stop presenting themselves as professional care. Until then, “AI therapy chatbot ban” is better understood as a political label than as a single description of what these laws do.
GPT-Live lets ChatGPT listen and speak at once, and OpenAI says it can bring in GPT-5.5 for harder work without stopping the conversation. The launch’s real test is whether that experience holds up across long sessions, regional capacity and uneven networks—and whether its related WebRTC proposals become broadly deployed infrastructure rather than a company-specific advantage.
Xero's survey of 1,081 New Zealand SME customers finds 61% are proactively using AI, but most users are learning by experimentation. Its new 12-week bootcamp with ASB is a response to that confidence gap; it is not yet evidence of a productivity gain.
Apple alleges that OpenAI and former Apple employees sought confidential hardware information through recruiting, supplier contacts and a former engineer’s post-employment system access. The complaint has not been tested in court, and OpenAI says it is unaware of evidence that it has merit.
Amazon’s market value passed $3 trillion after AWS revenue accelerated and the company raised its 2026 capital-spending plan to $220 billion. The rally reflects confidence in Amazon’s cloud demand, but the company’s own capacity forecast, the broad uses of its budget and strong rival cloud growth leave the return on that spending unproven.
The Trump administration says it has completed a voluntary framework for testing the cyber capabilities of the most advanced U.S. AI models and will discuss it with major developers Tuesday. The key scope, testing and disclosure rules have not been made public.
MediaTek now says its second custom cloud AI accelerator is on track for high-volume production in 2028, later than its April goal of entering production by the end of 2027. The company is betting that supply-chain coordination and advanced packaging can support a much larger cloud-chip business, but it has not disclosed a customer, package supplier, cost comparison or 2028 revenue guide.
A Nvidia stockholder’s derivative complaint seeks to make directors and senior officers answer for alleged copyright and biometric-data violations in the company’s AI work. The immediate question is not whether those underlying allegations are true, but whether the suit can clear the procedural and factual hurdles for a board-level claim.
Dongfang Suanxin says its planned 14nm DF2000 will use stacked near-memory computing to reach 15 TB/s of bandwidth. The company’s system-level comparison with NVIDIA highlights an architectural bet, but it does not establish a performance win or answer the questions of yield, cost and software adoption.
Mexico’s exports of computer equipment have surged as U.S. data-center construction and tariff differences reshape trade. The figures point to a bigger role in assembling and shipping equipment, but imported components, near-full factories and limited investment leave the higher-value parts of the chain elsewhere.
Bayreuth Festival says its anniversary Ring uses AI to recombine 150 years of Wagner imagery into a different visual experience at each performance. Early reviews and boos after Götterdämmerung do not settle the experiment, but they focus attention on the people choosing and deploying the system when its projections compete with singers and drama.
A Block technical analysis says a firmware path used to create some COLDCARD secrets could be reproducible; reported address sweeps make the risk urgent, but the analysis does not establish that every affected wallet can be recovered remotely.
Honolulu’s state-run airport has added 17 AI-generated songs to its hourly sound system. The available reporting does not show that they replaced the licensed music already played there, but it has prompted local curator Roger Bong to argue that a public gateway should be more transparent about who decides what Hawaiʻi sounds like.
Trump Media's Truth API offers licensed, machine-readable delivery of selected Truth Social posts to institutional customers. Its commercial appeal rests on the speed with which traders receive communications from President Donald Trump and other officials, but the available accounts leave unresolved exactly what advantage buyers receive.
OpenAI and Anthropic both support disclosure, incident reporting and outside scrutiny for the largest frontier-AI developers. Their proposals—and Illinois’ new law—show that the more consequential unresolved question is whether an evaluator or government agency can decide when a model’s risk is too high for deployment.
AMD’s Kria AI system-on-module and robotics developer platform combine an X100 processor, FPGA-equipped carrier board and open software stack. But the headline 3.4x real-time result comes from an AMD-commissioned simulation run on a Strix Halo mini PC configured as an X100 proxy—not on the forthcoming Kria hardware—and its public descriptions contain methodological differences that make independent reproduction the next test.
A Manhattan federal judge allowed Reddit’s core DMCA and conspiracy claims against Perplexity and SerpApi to proceed over alleged scraping through Google results. The ruling accepts a plausible theory tied to Reddit’s Google license, but leaves unresolved whether Reddit can prove authorization, protected works and actual circumvention.
MediaTek says its AI ASIC business could contribute about $2 billion in the fourth quarter of 2026. That is a company target for one unnamed US hyperscaler project, distinct from its longer-term market-share goal and a separate report on revenue mix.
OpenAI says it closed a $122 billion funding round at an $852 billion post-money valuation. Amazon’s filing sets out three linked elements: $15 billion already invested in OpenAI, a $35 billion share-purchase commitment, and an AWS commercial commitment expanded by $100 billion over eight years.
Two House committee chairs have asked DoorDash to identify the Chinese AI models it uses and the security testing behind them. DoorDash’s public benchmark shows Kimi K2.6 in one experimental code-review configuration, while its stated production reviewer used Claude models—leaving deployment scope and data controls unresolved.
Hank Green says he will reduce or pause work on several YouTube channels after acknowledging that he had leaned too heavily on ChatGPT-generated research notes. He maintains that the disputed line in an Ask Hank Anything episode was an ad-lib, not chatbot text; the unanswered question is how an education creator will make its AI-assisted research process legible to viewers.