AI Daily Digest
Your daily briefing on AI developments that matter to your business.
🎧 Listen to this digestThursday, November 19, 2026
Government & Policy
3Major Banks Declare AI Agent Hack Losses Are Now the Customer's Problem
Citi, PNC, and Citizens Bank have updated their terms of service to clarify that losses resulting from AI agent hacks — where autonomous AI systems authorized by customers are compromised — fall on the customer, not the bank. This creates a new liability gap as AI agents increasingly execute financial transactions on behalf of users. The policy shift comes as AI agent adoption accelerates and security researchers warn that agent-layer attacks are a growing threat vector.
AI WeeklyFederal Register Calls for Comments on Use of 'Generative Super Intelligence' in Agency Adjudication
The federal government has published a formal request for comments in the Federal Register on policies governing the use of generative AI — now officially termed 'Super Intelligence' — in agency adjudication processes. This follows the Trump administration's rebranding of AI across federal agencies and signals that formal rulemaking is beginning, at least for high-stakes government decision-making contexts. The NCUA and Treasury's AI Executive Oversight Group are also developing practical cybersecurity and risk management tools for financial institutions using SI.
Federal RegisterGovernor Newsom Pitches California as National AI Safety Leader Ahead of Likely 2028 Presidential Run
California Governor Gavin Newsom is publicly positioning himself as the counterweight to the Trump administration's hands-off AI approach, emphasizing California's independent AI safety framework and enforcement infrastructure. The move is widely read as an early signal for a 2028 presidential campaign, with Newsom framing AI regulation as a voter-anxiety issue where federal inaction creates a leadership vacuum. This comes days after California strengthened its auditor registry and verification standards for high-risk AI systems.
San Francisco ChronicleMajor AI Players
2Google Launches Gemini 4 Neon, a Mid-Tier Reasoning Model That Tops Benchmarks Against Claude and GPT
Google has released Gemini 4 Neon, a reasoning model positioned between its Hydrogen (fast) and Carbon (heavy) tiers. At launch, Neon ranked third overall on the Artificial Analysis Intelligence Index but led competitors on AutomationBench (56.3% vs. Claude Opus 5.5's 42.5%) and SWE-sweep, outperforming both Claude Opus 5.5 and GPT-6.2 Sol on several key coding and automation benchmarks. It is designed for mid-complexity workloads like technical review and iterative coding tasks where Carbon's full power is unnecessary.
Analytics VidhyaGoogle DeepMind's AlphaProof Code Uses Multi-Agent System to Formally Verify Signal Protocol in Four Weeks
Google DeepMind introduced AlphaProof Code, a multi-LLM agentic system in which Gemini Carbon agents generated over 6 million lines of Lean code to formally verify Signal's cryptographic protocol — a task that would have taken years manually. Agents flagged only about 1% of code for human review while the Lean proof checker verified the rest automatically. The project demonstrates that AI agents can handle large-scale formal verification tasks that were previously impractical.
AI WeeklyGlobal Developments
1AI Engineering Market Forecast to Reach $365B by 2036, Driven by Agentic AI and Asia-Pacific Adoption
A new market analysis projects the global AI engineering market will reach $365.1 billion by 2036, with Asia-Pacific registering the highest growth rate driven by enterprise adoption, national AI initiatives, and India's large delivery capacity. Europe's share is boosted by EU AI Act compliance demand across documentation, evaluation, and monitoring. The Middle East — led by Saudi Arabia and the UAE — is investing in sovereign AI, while Latin America sees Brazil lead with Mexico and Colombia expanding nearshore AI delivery.
GlobeNewswireRobotics
2Toyota Plans 400,000-Robot Factory Automation Program at $6.4B Annual Spend from 2028
Toyota is preparing what may be the largest factory automation program in automotive history, targeting approximately 150,000 robots inside its own factories and roughly 250,000 more across group companies and suppliers. Annual spending could reach $6.4 billion from 2028, with the fleet mixing traditional industrial arms with AI-driven machines including ELEY, a learning robot that observes skilled workers and replicates their movements. Toyota frames the initiative as human-robot partnership, with robots absorbing repetitive and physically demanding tasks while experienced workers guide operations.
Discovery Hub / FacebookHumanoid Robot Market Reality Check: Only 7,000 Units Sold in 2025 Against 600,000 Industrial Robots
Despite heavy media attention, the global humanoid robot market remains nascent — the International Federation of Robotics estimates just 7,000 full-size humanoids were sold worldwide in 2025, compared to over 600,000 traditional industrial robots installed in factories the same year. Current humanoids still lack the dexterity and adaptability needed for complex tasks like SKU picking or assembly lines, and cost several times more than task-specific robots while delivering lower throughput. Gartner and IFR both cite safety standards, high maintenance costs, and weak ROI cases as barriers to wider industrial adoption.
SupplyChainBrainTransportation
1Tesla Cybercab Fleet Surpasses Waymo in Texas Registrations with 1,507 Vehicles, Expands to Eight Cities
Tesla's Cybercab fleet has grown to 1,507 vehicles registered for driverless operation in Texas, overtaking Waymo's 1,321 — a lead Tesla first established on November 10 before Waymo briefly reclaimed it. Unsupervised autonomous service is now live in Austin, Dallas, Fort Worth, Houston, San Antonio, Miami, Orlando, and Tampa, with Austin operating 24/7 across roughly 400–440 square miles. Tesla is also planning to integrate Cybercabs into its service loaner program, replacing Uber credits and loaner cars at service locations.
TeslaratiHealthcare & Biotech
1Utah's AI Prescription Pilots Expose FDA Oversight Gap in Federal Clinical AI Roadmap
Utah is running active pilots where AI systems can renew or issue prescriptions for a few dollars, with clinician oversight designed to decrease over time — but the FDA has not stated whether these tools require premarket authorization. The HHS Clinical AI Roadmap, published last week, left this question unanswered, creating an accountability gap for patients who cannot determine which agency oversees the AI making their healthcare decisions. Third-party reviews have flagged potential vulnerabilities under deliberate misuse, though no serious safety incidents have been reported so far.
Medical DailyInvestment Noteworthy
1Anthropic IPO Prices at $83, Valuing Company at ~$1.5T — Largest IPO Ever, Delivering $100B+ to Venture Backers
Anthropic shares were priced at $83 on Wednesday evening, valuing the company at approximately $1.5 trillion on a fully diluted basis, making it the largest IPO in history. A small group of early venture investors — including Iconiq Capital (~$28B stake value), Lightspeed Ventures (~$25B), and Menlo Ventures (~$21B from a $1B bet) — are set to collect a combined windfall exceeding $100 billion. The IPO is being called a watershed moment for venture capital, validating the strategy of making outsized bets on AI frontier labs.
The Business TimesGeneral AI
1New York Breaks Ground on Empire AI Supercomputing Center, Launching Statewide AI Research Ecosystem
New York State has broken ground on the Empire AI Supercomputing Center, a shared research infrastructure being used by Columbia, Mount Sinai, and other universities for projects ranging from NYC traffic safety to mental health prediction and drug discovery. The launch follows Governor Hochul's creation of the New York Center for AI Responsibility and Research at SUNY Binghamton, bringing together more than 80 scientists to independently evaluate AI safety and performance. Together, the initiatives represent a major state-level investment in both AI capability and AI governance infrastructure.
SUNY