AI News Roundup – Nvidia CEO ousted amid infidelity and espionage scandal, Google DeepMind announces new Gemini models, UK to leverage National Health Service data for pharma AI goals, and more
- November 16, 2026
- Snippets
Practices & Technologies
Artificial IntelligenceTo help you stay on top of the latest news, our AI practice group has compiled a roundup of the developments we are following.
- Nvidia CEO Jensen Huang has resigned amid a growing infidelity scandal, with one mistress allegedly having ties to Chinese intelligence services, according to Bloomberg. Huang had denied allegations of marital infidelity internally since late 2025, and he initially resisted stepping down after reports surfaced late last month linking him to Alice Huang, the leader of a Singaporean AI company (and Nvidia customer) suspected of smuggling AI chips into China. Alice Huang had been suspected by U.S. intelligence officials of having links to Chinese foreign intelligence agencies. The scandal erupted earlier this month when the Financial Times reported the relationship between the two was romantic, prompting Nvidia general counsel Tim Teter to confront the CEO at his home and secure a confession. Nvidia’s board of directors subsequently hired the law firm Wachtell, Lipton, Rosen & Katz to investigate whether Huang had been targeted by a foreign intelligence operation, while Huang apologized for his actions in a company-wide email. This past week, the New York Times revealed that Huang had also engaged in an extramarital affair with Kimberly Powell, Nvidia’s vice president of health care, who reported directly to him. Following an emergency remote board meeting that stretched late into the night on November 11, Huang resigned as CEO and a member of the board, ending a storied career at the chipmaker he led since co-founding the company in 1993. After the resignation was announced, Reuters reported that Huang had also been involved in an intimate relationship with a second unnamed Nvidia employee. Nvidia’s board named Chief Financial Officer Colette Kress as interim CEO on Thursday morning, though Huang’s fall from grace has roiled markets. Nvidia shares dropped 5.5% on Thursday, bringing total declines to more than 30% since their October peak, as the board searches for a permanent replacement. Huang is preparing to testify before Congress regarding his interactions with Alice Huang as lawmakers investigate potential security risks to the company’s chip technology.
- Google DeepMind has released Gemini 4 Hydrogen, its newest “workhorse” AI model, as well as other new AI products. Replacing Gemini 3.8 Flash as a default model across Google’s AI ecosystem, Gemini 4 Hydrogen powers an upgraded version of Google Search and costs developers $1.50 per million input tokens and $7.50 per million output tokens, maintaining the same price level as its predecessor. Operating across four adjustable reasoning levels, Hydrogen runs 2.5 times faster than its predecessor, according to the company, cuts fabricated claims (or so-called hallucinations) by more than fivefold, and rivals higher-tier models such as Gemini 4 Argon in coding benchmarks. The Hydrogen model is now available to Google’s AI Pro and AI Ultra subscribers, as well as certain enterprise customers. Separately, WIRED reports that Google DeepMind used a new agentic framework called AlphaProof Code to mathematically prove, or formally verify, the security of the Signal encryption protocol that protects messaging apps such as the eponymous Signal and Meta’s WhatsApp. According to a company blog post, the multi-agent system used models powered by Gemini 4 Carbon (an advanced variant in the Gemini 4 family) to generate more than 6 million lines of verification code in the Lean programming language. The process then automatically audited its own proofs so that DeepMind’s human experts needed to review only 1% of the codebase, and it accurately flagged two historical vulnerabilities when tested on older versions of the Signal protocol. DeepMind has open-sourced the verification proofs to help improve future AI-assisted cybersecurity testing, though some researchers cautioned that scaling up such automated formal software verification will likely remain a challenge, especially amid rising costs of deploying AI models.
- A U.K. government minister proposed using health data from the country’s National Health Service to aid pharmaceutical innovation through AI, according to the Financial Times. Speaking at the FT’s Global Pharma and Biotech Summit, U.K. science and health minister Chris McDonald argued that combining Britain’s decades-long longitudinal medical records from the NHS with the country’s AI sector gives the nation a strategic competitive advantage in drug discovery. McDonald pledged to continue the data commercialization effort started under former Prime Minister Keir Starmer into the current administration of Andy Burnham, using the offer of NHS datasets to attract pharmaceutical investment and support drug-discovery-focused AI firms such as the London-based Isomorphic Labs. To encourage drugmakers to invest domestically, the government recently raised NHS prices for new medicines under a deal with the U.S. government, a policy projected to cost the health service £1 billion over the next three years while generating broader tax revenue and accelerating the development of new treatments. However, the initiative faces public skepticism over patient privacy and the role of the U.S. contractor Palantir, whose health service contract is being extended by two years while officials prepare a fresh procurement process focused on domestic companies. McDonald stressed that “the only way we can exploit [the NHS data] is if we maintain the trust and confidence of society.”
- The New York Times reports that a so-called AI law firm has acquired legal malpractice insurance for its AI agents. New York-based Crosby, which employs about 60 attorneys to oversee automated contract reviews, spent six months securing a $5 million liability policy from Testudo, a specialty insurance startup backed by Lloyd’s of London. Crosby founder Ryan Daniels sought coverage for autonomous tasks that fall into a gray area between clerical work and legal practice, such as answering questions about a contract. Once the policy took effect, Crosby removed attorney oversight from those gray-area tasks, which account for roughly 20% of the firm’s workload, though lawyers still supervise strictly legal work to comply with state regulations barring the unauthorized practice of law. The novel arrangement highlights a growing coverage gap in the insurance market, as at least 16 of the 20 largest liability carriers have sought regulatory approval to exclude generative AI from policies amid rising hallucination claims and uncertain risks surrounding rapidly evolving AI tools. Startup insurers such as Testudo are emerging to underwrite these risks by auditing company guardrails and profiling litigation exposure, seeing an untapped market as AI use continues to rise in the legal services industry.
- United Airlines is banning AI agents from booking flights for its customers, according to the Wall Street Journal. While personal assistants such as Meta’s Muse and OpenAI’s Dots can still search flight schedules and fares, the airline is now blocking them from completing ticket purchases on its website. Airline executives warned that third-party bots could expose sensitive passenger data, make unintended purchases, or violate federal aviation rules requiring travelers to personally confirm safety disclosures, such as attesting that luggage contains no explosives. United CEO Scott Kirby also expressed reluctance to empower a new digital middleman that could disrupt direct customer relationships and advertising, echoing past airline disputes with online travel agencies. United’s approach to AI agents contrasts with rival airlines such as Delta and American Airlines, which permit authenticated agents to purchase tickets under certain safeguards. Instead, it aligns United with others that have blocked agents from accessing their platforms, most notably e-commerce giant Amazon, which blocked Meta’s Muse in September. The decision reflects wider disputes about AI agents across the financial and retail sectors; as the Journal notes, banks and merchants are scrambling to manage a flood of autonomous shopping bots accessing their platforms. While payment companies such as Stripe are rolling out tokenized wallets and single-use virtual cards to enable bot transactions, major lenders such as JPMorgan Chase are blocking agent access over elevated fraud, return rates, and regulatory scrutiny, as this roundup reported in October.


