New York Attorney General Letitia James has created the AI Safety, Standards, Enforcement, and Transparency Unit, or ASSET, to enforce AI safety laws and address consumer privacy, cybersecurity, digital scams and online safety. Led by Cristy Phillips, it will sit in the Economic Justice Division and work with the Bureau of Internet and Technology. ASSET will oversee the Responsible AI Safety Education Act, effective Jan. 1, 2027, which requires developers of certain advanced AI models to disclose safety measures and promptly report safety or security incidents. The unit will also support investigations and develop policy proposals.
AI infrastructure demand remains strong, but financing and capacity risks loom. CoreWeave reported third-quarter revenue growth of 168% year over year and raised its 2026 revenue forecast to $13.0 billion–$13.3 billion. Its backlog climbed 138% to about $132.5 billion, and CEO Michael Intrator said the company is largely sold out through 2027, with qualified demand roughly 10 times its available capacity through the end of next year. OpenAI, Perplexity and Jane Street expanded their relationships with CoreWeave, which expects about $9 billion in customer prepayments and deposits over the next three months. Yet quarterly net interest expense reached $846 million—more than twice adjusted operating income—and is expected to rise to $1.05 billion–$1.15 billion in the fourth quarter. CoreWeave says it will avoid new unsecured parent-company debt through 2027 and expects contract cash flow to cover repayments. Analysts warn that easing capacity shortages could hurt the company; Nvidia’s upcoming earnings are another gauge of AI demand.
Jet.AI is also moving toward AI infrastructure, reporting a third-quarter net loss of about $5.9 million and a $1.75 million investment for a 15% stake in StratGrid and its Wheatland Project in Alberta. The project aims to generate 240 MW for off-grid data centers. Regulatory submission is targeted for early 2027, with construction potentially beginning in the second half of that year. Separately, a report on AI investment describes rising debt and spending exposure: the top 10 AI-exposed companies account for about 40% of the U.S. stock market. Anthropic, which has more than $130 billion in cash reserves ahead of a planned listing, says it may invest up to $900 billion over five to 10 years. Goldman Sachs forecasts Anthropic revenue of $53 billion this year and as much as $636 billion by decade’s end; the article characterizes its planned stock float as unusually risky.
Companies are adding controls for AI agents while preparing their documentation and workforces for wider adoption. CrowdStrike’s Falcon Guardian now includes an AI Kill Switch to stop unauthorised agents and block them from running again across an organisation. Its generally available AI Gateway inspects prompts and responses, monitors use and applies policies on managed and unmanaged devices. Agent-level hooks support Claude Code, OpenAI Codex, GitHub Copilot and Cursor; Windows endpoint network inspection is available, with Mac support in early access. CrowdStrike also plans to acquire Fiddler, subject to closing conditions, for agent-observability technology that could help assess agents before deployment. At Revolut, almost seven years of design documentation became AI-readable in one day, giving the team time to review and fix template and wording problems before using the material to support a new designer tool.
AI spending is also feeding into economic and workforce debates. On Nov. 17, 2026, St. Louis Fed President Musalem said data-center investment and related demand for semiconductors, memory chips, construction and energy are adding to prices, while promised productivity gains remain unproven. He argues rates of 3.75%–4% are about one percentage point too low and has called for several hikes by spring 2027; he cannot vote on Federal Open Market Committee decisions until 2028. Meanwhile, demand for junior technology workers with AI skills has risen fivefold since early 2025, and 86% of leaders plan to expand technology teams. Demand for AI risk-governance skills rose 108%, against a 69% increase in supply. Three in five employers say human skills such as critical thinking and communication are harder to find than technical skills; two-thirds prioritize flexible workers and 92% use digital talent marketplaces. Women account for 31% of professionals with emerging AI skills, versus 36% of workers in older technology roles.
Model testing points to a role for routing tasks rather than relying on one system. Fireworks AI says MiniMax M3.1 Flash completed more tasks than an unnamed comparison model on the DeepSWE coding test and correctly refused a request for a higher credit limit in a banking test, though it added a statement credit to soften the refusal. The system can send DeepSWE and On Escalation tasks to Flash first and pass only failed tasks onward. M3.1 Flash costs less than Gemini for coding, while the article says banking and research work favor routing among models rather than replacing them.
Key Takeaways
- New York’s ASSET unit will enforce AI safety laws and oversee the Responsible AI Safety Education Act, which takes effect Jan. 1, 2027.
- CoreWeave’s third-quarter revenue rose 168%; its backlog reached about $132.5 billion, but quarterly net interest expense was $846 million and is expected to increase.
- CoreWeave says it is largely sold out through 2027 and expects about $9 billion in customer prepayments and deposits over the next three months; OpenAI, Perplexity and Jane Street expanded ties with the company.
- Jet.AI invested $1.75 million for a 15% stake in StratGrid’s Alberta project, which targets 240 MW for off-grid data centers.
- Anthropic has more than $130 billion in cash reserves and says it may invest up to $900 billion over five to 10 years; Goldman Sachs forecasts revenue could reach $636 billion by decade’s end.
- CrowdStrike’s AI Kill Switch can stop unauthorised agents; its agent hooks support Claude Code, OpenAI Codex, GitHub Copilot and Cursor.
- Revolut converted nearly seven years of design documentation into AI-readable material in one day, then reviewed and corrected issues before using it for a new designer tool.
- St. Louis Fed President Musalem says AI infrastructure spending is adding to inflation, and argues rates should be about one percentage point higher; he cannot vote on rate decisions until 2028.
- AI-skill demand and hiring plans are rising: junior-worker demand has increased fivefold since early 2025, while demand for AI risk-governance skills rose 108%.
- Fireworks AI says MiniMax M3.1 Flash performed well on a coding test and refused a higher credit limit in a banking test; it costs less than Gemini for coding, while task routing remains useful for banking and research.
New York creates unit to oversee AI safety and enforcement
New York Attorney General Letitia James announced the AI Safety, Standards, Enforcement, and Transparency Unit, known as ASSET, to protect people from risks linked to artificial intelligence and other new technologies. The unit will enforce AI safety laws, protect consumer privacy, and address cybersecurity, digital scams, and online safety. It will oversee the Responsible AI Safety Education Act, which takes effect Jan. 1, 2027, and requires developers of certain advanced AI models to disclose safety measures and quickly report safety or security incidents. The unit will also develop policy proposals and provide technical support for investigations. Cristy Phillips will lead it.
New York launches ASSET unit to police AI risks
New York Attorney General Letitia James announced a new unit called the AI Safety, Standards, Enforcement, and Transparency Unit, or ASSET. It will enforce AI safety laws, protect consumer privacy, and address cybersecurity, digital scams, and online safety concerns. The unit will oversee New York’s Responsible AI Safety Education Act, which takes effect Jan. 1, 2027, and will develop policy proposals and provide technical help for investigations. It will work within the Attorney General’s Economic Justice Division alongside the Bureau of Internet and Technology. Cristy Phillips, a former special counsel for the division, will lead the unit.
New York attorney general forms unit to enforce AI safety laws
New York Attorney General Letitia James announced the AI Safety, Standards, Enforcement, and Transparency Unit, known as ASSET, to address risks from artificial intelligence and other emerging technologies. It will protect consumer privacy, address cybersecurity threats and digital scams, and oversee enforcement of the Responsible AI Safety Education Act, which takes effect Jan. 1, 2027. The law requires developers of certain advanced AI models to disclose safety measures and promptly report security and safety incidents. The unit will support investigations and develop policy proposals, while working within the Economic Justice Division and with the Bureau of Internet and Technology. Cristy Phillips will lead the unit.
CoreWeave results show strong AI demand but high financing costs
CoreWeave’s third-quarter revenue rose 168% from a year earlier and beat analyst expectations. The company raised its 2026 revenue forecast to $13.0 billion to $13.3 billion, from a previous range of $12.4 billion to $13.2 billion. The results eased worries about whether customers still want AI computing, though the article notes that financing the data center buildout remains a major concern. CoreWeave reported $846 million in net interest expense for the quarter, more than twice its adjusted operating income, and expects $1.05 billion to $1.15 billion in the fourth quarter. Analysts also warned the company could be hit hard if the shortage of AI computing capacity eases.
CoreWeave demand surges as Nvidia prepares to report earnings
CoreWeave shares rose about 14% in premarket trading after the company reported third-quarter results on Nov. 10. Revenue increased 168% year over year, and its revenue backlog grew 138% to about $132.5 billion. CEO Michael Intrator said CoreWeave was largely sold out through 2027, with about 10 times more qualified demand than available capacity through the end of next year. OpenAI, Perplexity, and Jane Street expanded their relationships with the company, which expects about $9 billion in customer prepayments and deposits over the next three months. CoreWeave also said it does not plan to issue more unsecured parent company debt through 2027 and expects cash flow from existing contracts to cover debt repayments, while Nvidia’s upcoming earnings report could provide further insight into AI demand.
Jet.AI invests $1.75 million in Alberta data center project
Jet.AI reported a net loss of about $5.9 million for the third quarter of 2026 and said it invested $1.75 million in StratGrid Inc. The investment gives Jet.AI a 15% interest in StratGrid and its Wheatland Project in Alberta, which aims to generate 240 MW of power for off-grid data centers. Jet.AI said it is shifting toward AI infrastructure after its merger with flyExclusive and is discussing other strategic opportunities. The Wheatland Project is on track for a regulatory submission in early 2027, with construction potentially starting in the second half of that year.
AI investment boom faces growing debt and spending risks
Investment in artificial intelligence is drawing huge amounts of debt and money, with the top 10 AI-exposed companies making up about 40 per cent of the American stock market. Anthropic and OpenAI are competing to build advanced AI systems, while Anthropic has more than $US130 billion in cash reserves ahead of a planned stock market listing. Its prospectus says it may invest up to $US900 billion over the next five to 10 years. Much of that spending is expected to come from revenue, and Goldman Sachs predicts Anthropic could earn $US53 billion this year and as much as $US636 billion by the end of the decade. The article describes Anthropic's planned stock float as unusually risky.
CrowdStrike adds AI kill switch for unauthorised agents
CrowdStrike introduced an AI Kill Switch in Falcon Guardian that can stop unauthorised AI agents and prevent them from running again across an organisation. Its AI Gateway is generally available and inspects prompts and responses, monitors use, and enforces policies on managed and unmanaged devices. Agent-level hooks support Claude Code, OpenAI Codex, GitHub Copilot and Cursor, while additional protections address prompt injection, sensitive personal information and malicious AI skills and MCP servers. Endpoint network inspection is available for Windows, with Mac support in early access. CrowdStrike also plans to acquire Fiddler, whose agent observability technology could help organisations assess agents before deployment, but the acquisition remains subject to closing conditions.
Revolut makes seven years of design documents AI-ready in one day
Revolut turned almost seven years of design documentation into AI-readable documents in one day, according to senior design system designer Dmytro Pavlov. The design system serves as a shared language for people outside and inside the company, and the documents will support a new AI tool for designers. Pavlov said the quick draft gave the team time to review the results and correct problems, including flaws in the document template and unclear wording in visual documents. The article argues that reviewing and improving documentation is important before using it to support AI tools.
Fed's Musalem says AI spending is adding to inflation now
On November 17, 2026, St. Louis Fed President Musalem said AI infrastructure spending is pushing up demand and prices, while promised productivity gains have not yet been proven. He linked inflation above the Fed’s 2% target to negative supply shocks and AI-driven demand, especially investment in data centers. Spending on semiconductors, memory chips, construction and energy has also raised prices. Musalem has called for several interest rate hikes by spring 2027, arguing that the current rate of 3.75% to 4% is about one percentage point too low to return inflation to target. He is not a voting member of the Federal Open Market Committee until 2028, so he cannot formally vote on rate decisions before then.
AI hiring shifts to buying, building and borrowing talent
Demand for junior technology workers with AI skills has grown fivefold since the start of 2025, and 86% of leaders plan to expand their technology teams. Employers are competing for workers, while AI micro-credentials are gaining recognition alongside university degrees. Demand for AI risk governance skills rose 108%, compared with a 69% increase in supply, and three in five employers say human skills such as critical thinking and communication are harder to find than technical skills. Women make up 31% of professionals with emerging AI skills, compared with 36% of workers in older technology roles. To get skills they need, two-thirds of employers are prioritizing flexible workers, and 92% use digital talent marketplaces.
MiniMax M3.1 Flash excels at coding but falls short in banking
Fireworks AI reported that MiniMax M3.1 Flash correctly denied a request for a higher credit limit in a banking test, then added a statement credit to soften the refusal. On the DeepSWE coding test, M3.1 Flash completed more tasks than an unnamed comparison model. The system can route DeepSWE and On Escalation tasks to Flash first, sending only failed tasks onward. M3.1 Flash costs less than Gemini for coding, but the article says banking and research tasks favor routing work among models rather than replacing them.
Sources
- NY Attorney General James announces new unit to oversee AI safety and enforcement
- NY Attorney General James announces new unit to oversee AI safety and enforcement
- NY Attorney General James announces new unit to oversee AI safety and enforcement
- CoreWeave Answered the Biggest Question About AI Stocks. Now Nvidia Has to Confirm It.
- CoreWeave Answered the Biggest Question About AI Stocks. Now Nvidia Has to Confirm It.
- Jet.AI Inc. Reports Third Quarter 2026 Results and Updates AI Infrastructure Strategy, StratGrid Investment and Reverse Takeover Process
- Why the wheels are falling off the AI investment boom
- AI Security: CrowdStrike Adds Kill Switch For Unauthorised Agents
- Revolut turned 7 years of docs AI-ready in a day
- Fed's Musalem says AI is pulling demand forward, and policy can't wait
- AI-Ready Workforce: Buy, Build, Borrow Strategies
- MiniMax Flash wins coding, loses banking
Comments
Please log in to post a comment.