Georgia Closes with Three AI Bills, NHS Sets Scribe Rules

2026-04-06

Conversational AI Watch

The news that moves policy, portfolios, and patient safety.

By Jess Jessop | April 6, 2026 | Issue #8

WATCH | QUICK LISTEN | DEEP DIVE | READ ON WEB

JESS'S TAKE

Georgia's legislature adjourned today with three AI bills sitting on the governor's desk. One of them, SB 540, requires chatbots to tell you they are chatbots. The fact that this needs a law tells you everything about 2026.

Meanwhile, across the Atlantic, NHS England just published national guidelines for AI ambient scribes. The core message: patients can say no.

That sentence is doing more work than most 40-page policy papers. Consent is not a checkbox. It is a conversation.

In Nebraska, a chatbot safety bill just hitched a ride on an agricultural privacy act. If that sounds strange, you have not been watching statehouses this year. Bipartisan momentum finds the vehicle that moves.

And Netsmart is making the only FDA-cleared autism assessment technology available to its entire provider network during Autism Acceptance Month. Eye-tracking. Children as young as 16 months. No waitlist required.

The throughline today is transparency. Who is talking to you. What is listening. What happens to your data. The states are not waiting for Washington. The NHS is not waiting for anyone.

THE SIGNAL / AI SAFETY AND REGULATION

Georgia Session Ends with Three AI Bills on Kemp's Desk

Georgia's legislature adjourned today after passing three AI-related bills in a session where lawmakers openly navigated tension with the Trump administration's December executive order threatening federal funds for states that regulate AI.

SB 540 requires companion chatbots to disclose their AI nature every three hours for adults and hourly for minors, with protocols for detecting suicidal ideation and self-harm. SB 444 bans insurance companies from making healthcare coverage decisions based solely on AI. SR 789 creates a study committee on AI's broader impact.

Rep. Todd Jones deliberately delayed SB 540's effective date to next summer, giving Congress roughly 15 months to act. The message to Washington: the states will wait, but not forever.

Takeaway: Georgia threaded the needle between bipartisan safety legislation and Trump administration deference. The delayed effective date is a template other red states may follow: pass the bill now, let Washington try to preempt it later.

For Legislators: The Georgia model shows how to advance chatbot safety without triggering the DOJ AI Litigation Task Force. Delayed effective dates and narrow scope give federal lawmakers a window to act while protecting constituents now.

Sources: Atlanta Journal-Constitution, Transparency Coalition AI Legislative Update

THE SIGNAL / DIGITAL HEALTH INNOVATION

NHS England Sets National Rules for AI Scribes: Patients Can Say No

NHS England published comprehensive guidance on AI-powered ambient scribing tools, developed with the Information Commissioner's Office and the National Data Guardian. The tools passively listen to clinical conversations and generate consultation notes, summaries, and letters.

The central rule: clinicians must tell patients at the start of every interaction if an AI scribe is running. Patients can object, and if they do, the tool must be turned off. Period.

A major NHS-sponsored study found AI scribes increased direct patient interaction time by 23.5% while reducing appointment length by 8.2%. The guidance also mandates that organizations using non-compliant AI scribe solutions face direct liability.

Takeaway: The NHS is building the governance framework that the U.S. still lacks. Ambient AI in clinical settings works, but only when transparency and patient choice are non-negotiable. This is the standard other health systems should benchmark against.

For Legislators: The NHS guidance offers a ready-made template for U.S. state health committees considering AI documentation rules. The key principles: informed transparency (not explicit consent), patient veto power, and organizational liability for non-compliant tools.

Sources: Digital Health (UK), NHS England

THE SIGNAL / PATIENT SAFETY AND ETHICS

Netsmart Partners with EarliPoint to Bring FDA-Cleared Autism Assessment to Scale

Netsmart, a leading behavioral health EHR provider, announced a partnership with EarliPoint Health to distribute the only FDA-cleared early autism assessment technology across its provider network. The EarliPoint System uses eye-tracking to measure visual attention patterns in children as young as 16 months.

The timing is deliberate: announced during World Autism Acceptance Month, the partnership addresses a diagnostic backlog where families routinely wait months for evaluation. Current CDC data estimates one in 31 children is identified with autism, yet average diagnosis still occurs between ages four and five.

EarliPoint recently expanded its cleared age range to children up to 8 years. Southeast Missouri State University's Autism Center became one of the first providers to implement the expanded range starting April 1.

Takeaway: This is what responsible AI deployment looks like: FDA-cleared technology, objective biomarkers, distributed through existing clinical infrastructure. The partnership model solves the last-mile problem that keeps breakthrough diagnostics from reaching the families who need them.

For Legislators: Early autism identification reduces long-term costs across Medicaid, education, and behavioral health systems. State legislators should examine whether their autism assessment reimbursement policies cover FDA-cleared AI-assisted tools like the EarliPoint System.

Sources: Netsmart, KZIM/KSIM

THE SIGNAL / AI SAFETY AND REGULATION

Nebraska Chatbot Safety Bill Hitches Ride on Agricultural Privacy Act

In a legislative maneuver that captures 2026 perfectly, Nebraska Sen. Bostar's AI chatbot safety bill LB 1185 has been attached to the popular Agricultural Data Privacy Act (LB 525). The combined package appears cleared for passage, with the session scheduled to adjourn April 17.

LB 1185 contains requirements similar to Oregon's recently enacted chatbot safety law, including disclosure obligations and safety protocols. Attaching it to an agricultural bill is strategic: rural data privacy carries bipartisan support that chatbot regulation alone might not generate.

Takeaway: The legislative vehicle matters as much as the policy. Chatbot safety bills that cannot move on their own are finding bipartisan hosts. Expect more of this bundling strategy in states where standalone AI regulation faces political headwinds.

For Legislators: The Nebraska model suggests that data privacy omnibus packages may be the most viable vehicle for AI safety provisions in states where standalone chatbot bills face committee resistance. Consider whether your state's pending privacy legislation could carry complementary AI provisions.

Source: Transparency Coalition AI Legislative Update

THE SIGNAL / CONVERSATIONAL AI TECHNOLOGY

FDA Raising the Bar on What Counts as a "Breakthrough" AI Device

A STAT News analysis of the FDA's Breakthrough Device Tracker reveals a shift in what the agency considers worthy of its priority review designation. Since 2016, the FDA has granted breakthrough status to more than 1,200 devices, authorizing 160 to enter the market.

The trend: AI tools that merely improve what a doctor already does are no longer enough. The FDA is increasingly reserving breakthrough designation for AI that solves problems physicians simply cannot, such as detecting multiple cancers from a single image or predicting mortality risk from heart failure.

This comes as Congress considers bipartisan legislation to give breakthrough devices automatic Medicare coverage for four years. A Stanford study found it takes an average of 5.7 years after FDA authorization for a breakthrough device to achieve even nominal Medicare coverage.

Takeaway: The FDA is signaling that incremental improvement is not a breakthrough. For AI health companies pursuing 510(k) or breakthrough pathways, the message is clear: demonstrate something a clinician cannot do alone, not just something they can do faster.

For Legislators: The 5.7-year gap between FDA authorization and Medicare coverage creates a "valley of death" for breakthrough medical devices. The bipartisan proposal for automatic four-year Medicare coverage would close this gap and accelerate patient access to validated AI diagnostics.

Sources: STAT News, STAT News (Medicare)

THE SIGNAL / MENTAL HEALTH POLICY

One Mind Accelerator Demo Day Approaches: 14 Mental Health Startups, Nearly $1B in Portfolio

The One Mind Accelerator's fourth cohort will present at Demo Day on April 15 in Menlo Park. The 14 startups were selected from a record applicant pool at less than 5% acceptance rate, spanning pre-seed through Series A stages.

The portfolio now includes 51 companies that have collectively raised nearly $1 billion and reached over 1.5 million people. This year's cohort includes Flow Neuroscience (the first FDA-approved at-home tDCS headset for depression), Birches Health (insurance-covered addiction care), and Forecast Bio (AI-powered preclinical brain mapping).

Opening week featured investor panels with ARCH Venture Partners, Khosla Ventures, and Satori Neuro alongside federal panels with ARPA-H, NIMH, and SAMHSA.

Takeaway: One Mind remains the clearest signal of where institutional money and policy attention are converging in mental health innovation. The cohort's emphasis on serious mental illness and precision psychiatry reflects a market moving beyond wellness apps toward clinical-grade solutions.

For Legislators: State mental health innovation funds and Medicaid waivers should be structured to accommodate the AI-driven diagnostics and therapeutics emerging from accelerators like One Mind. The gap between innovation velocity and reimbursement policy remains the primary barrier to scale.

Source: One Mind

WHAT WE BUILT

Casey AI: The Therapist in the Loop Standard

Casey is a voice-first AI-native mental health EHR with Casey Life and Peer AI Coach supervised by licensed therapists. Every interaction is guided by clinician oversight. Every response has been tested across 1.78 million safety validations with 100% accuracy. Casey is on the FDA 510(k) pathway for pre-FDA safety validation as a Class II Software as a Medical Device.

The data layer features the first HIPAA-compliant Neo4j Memory Graph, which builds persistent therapeutic context across months of daily sessions. Casey uses they/the student/the client. No unsupervised AI. No unmonitored conversations. The therapist is always in the loop.

Watch the Demo →