ma2tic

The brief · 30 March 2026

The AI brief, 2026-03-30

Six vendors, one shared momentum: AI is moving into places where money, concrete, and mental health were still being managed the old-fashioned way. The question is who actually puts the guardrails in place.

3 min read J / K to navigate

Your AI clinical assistant is already making ethical calls for you

A study in npj Digital Medicine shows adaptive AI embedded in hospital software quietly resolves value trade-offs on its own. Sensitivity versus specificity, speed versus caution: these calls get baked into the code before the physician even opens the chart. Clinicians inherit an already-slanted decision without knowing it.

practitioners › Study published in npj Digital Medicine, vol. 9, March 30, 2026.

www.nature.com →

A survey finally puts numbers on how many Americans talk to AI about their mental health

You probably suspected people were venting to ChatGPT instead of seeing a therapist. This JMIR study puts figures behind that hunch, and more importantly it asks people what they actually think about it, not just whether they use it. What's still unclear is whether this replaces real care or just fills the gap while waiting three months for an appointment.

practitioners › Cross sectional survey published March 30, 2026 in JMIR Mental Health, examining use and perceptions of generative AI and chatbots for mental health support among US adults.

mental.jmir.org →

Glia launches a copilot that answers for your bank staff

Your bank advisor is increasingly leaning on AI to answer your questions, often without you noticing. Glia claims it can automate up to 80% of customer interactions, which mainly changes the speed and consistency of the answers you get at the counter or in chat.

practitioners › Glia CoPilot pairs a customer-interaction agent with an internal search tool, announced March 30, 2026.

Glia →

California Moves to Vet Which AI Vendors Can Sell to the State

Newsom just signed an executive order forcing California agencies to build their own certification criteria for AI vendors. If you're selling an AI product to the state, you'll need to prove your tool doesn't generate illegal content, doesn't discriminate, and protects privacy. This doesn't touch consumers directly yet, but it's setting a template other states are likely to copy.

practitioners › Executive Order N-5-26, signed March 30, 2026: agencies must produce certification standards, procurement guidelines, and security requirements (illegal content, bias, civil rights, privacy) before any contract goes through.

www.gov.ca.gov →