AINews Dispatch
Fresh technical briefings on the model race, regulation, and product shifts shaping AI right now.
Showing 109-120 of 256 briefings
A first-in-the-nation New York law now forces ads to admit when the person on screen isn't real. It's a small rule with a big tell: in the absence of a federal AI law, the states are writing the rulebook one statute at a time.
Read the briefingThe UK's AI Hardware Plan uses an Advanced Market Commitment to make government the first customer for inference-chip startups, even as private data-center spending dwarfs the public package.
Read the briefingSelf-driving laboratories are moving AI for science from passive prediction into closed-loop experimental systems that design, run, observe, and revise real-world tests.
Read the briefingA new Nature Medicine benchmark challenges dedicated clinical AI tools on answer quality, pushing healthcare AI's durable moat toward workflow, validation, auditability, and governance.
Read the briefingOpenAI's agreement to acquire Ona points Codex toward persistent, customer-controlled enterprise cloud runtimes with scoped credentials, audit logs, network boundaries, and review gates.
Read the briefingAnthropic's TCS partnership shows frontier AI moving from model access into systems-integration, governance, training, and workflow packaging for large regulated organizations.
Read the briefingVisa's OpenAI collaboration treats ChatGPT shopping as bounded payment infrastructure: tokenized credentials, authorization, agent identification, fraud monitoring, user limits, and approval flows.
Read the briefingCMS's new health-technology office is a sign that healthcare AI is moving from pilots into the machinery of procurement, interoperability, data exchange, and accountability.
Read the briefingAI safety is no longer only a lab governance debate. Courts, attorneys general, cybercrime disruption, telecom networks, the FBI, and G7 diplomacy are now turning model risk into public process.
Read the briefingAnthropic wanted government to have a way to stop dangerous frontier-model deployments. The Fable 5 and Mythos 5 shutdown shows how difficult that idea becomes when the process is opaque, urgent, and built around model access.
Read the briefingAs AI agents move from chat windows into enterprise workflows, security teams are starting to ask where all these software actors live, what they can touch, and who is accountable when they act.
Read the briefingGlean's Work AI Index gives a name to the hidden supervision labor around workplace AI: workers report large time savings, but also spend hours checking, correcting, and repairing AI output.
Read the briefing