AI Business | Jun 27, 2026

Analysis

• US AI policy tightened on two fronts in a single day: the White House asked OpenAI to limit its next model to government-approved users — after blocking foreign nationals from its systems and signing an order mandating nominally voluntary release reviews — while Reps. Josh Gottheimer and John Moolenaar introduced a bipartisan bill letting US cloud providers flag suspected foreign misuse of advanced AI compute to the government, converging executive and legislative pressure toward an AI-security control regime.
• OpenAI kept expanding even as Washington constrained it, naming former Uber India chief Prabhjeet Singh as its first India managing director for what it calls its second-largest market and detailing Jalapeño, a Broadcom-built custom inference chip to hedge Nvidia dependence — moves that land against the White House model-release limit and reported deliberations to delay its IPO.
• Cost discipline was the connective theme across research: AI startup Lindy moved 100% of its traffic off Anthropic's Claude to China's DeepSeek and expects to save millions, and researchers at the National University of Singapore unveiled MRAgent, an agent-memory framework claiming up to 27x lower token use — both signs of enterprises shifting from "tokenmaxxing" to efficiency, the same pressure pushing labs toward custom silicon.
• DeepSeek is converting its efficiency-driven breakout into a commercial push — a recruitment drive to "double" many core teams ahead of its first outside-investment raise — even as its consumer chatbot cedes ground to ByteDance's Doubao, underscoring how the open-source R1 cost advantage that matched OpenAI's GPT-4 and o1 is still reshaping the US-China map.
• Anthropic's "stay at the frontier to stay safe" posture — a near-$1 trillion valuation and US military customers — now runs directly into the administration it is fighting in a legal tussle over the military's use of its models, sharpening the tension between its safety mission and its accumulation of capital, compute and political influence.

POLICY & REGULATION

• LEGSemafor | White House limits OpenAI model release | Semafor reported the White House asked OpenAI to limit the release of its next model to government-approved users, cementing the Trump administration's shift toward AI interventionism. After starting laissez-faire and removing Biden-era safety-review rules, the administration has reversed course — entering a legal tussle with Anthropic over the military's use of its AI, blocking foreign nationals from accessing cutting-edge systems, and signing an order this month imposing nominally voluntary reviews of new releases. Politico described the result as an "open-ended and confusing regulatory landscape," and The New York Times reported OpenAI is considering delaying its IPO amid stock-market choppiness.
• LEGAxios | Gottheimer, Moolenaar roll out AI cloud security bill | Axios reported Reps. Josh Gottheimer (D-N.J.) and John Moolenaar (R-Mich.) are introducing legislation that would let US cloud companies flag suspected foreign misuse of advanced AI computing to the government, per a copy of the bill seen by Axios. The bipartisan measure reflects growing Capitol Hill interest in AI-security issues as the Trump administration works out its AI policy and frontier models advance.

PARTNERSHIPS & ENTERPRISE

• ENTTechCrunch | OpenAI poaches Uber India chief to lead its biggest market outside the U.S. | TechCrunch reported OpenAI appointed former Uber India and South Asia president Prabhjeet Singh as its first managing director for India, which it calls its second-largest market after the US. Singh, who resigned from Uber on Friday, joins in September and reports to Asia-Pacific managing director Kiran Mani, overseeing consumer growth, enterprise adoption, partnerships and regulatory engagement. The hire follows OpenAI's first New Delhi office last August, planned offices in Mumbai and Bengaluru, and partnerships with Indian conglomerates Reliance and Tata Group.
• ENTFinancial Times | DeepSeek plans hiring spree in escalation of China's AI talent war | Financial Times reported Chinese AI group DeepSeek plans to double the size of many core teams as it seeks to commercialise its frontier research, launching a recruitment drive on Friday to "expand every department." The openings — AI product managers, infrastructure engineers for AI computing clusters and training frameworks, and specialists in law, medicine and languages — signal founder Liang Wenfeng's ambition to build a broader AI company ahead of taking outside investment for the first time, even as DeepSeek's consumer chatbot cedes ground to ByteDance's Doubao.

GOVERNANCE & SAFETY

• SAFEWired | Anthropic Thinks Its Own Success Is Key to Making AI Safe | Wired reported that Anthropic, recently valued at almost $1 trillion and courting customers including the US military, operates on the belief that the world is better off with it at the frontier of the AI race. Former employees told Wired that leaders internally describe themselves as the "good guys" and treat the accumulation of capital, compute, research talent and political influence not as an end but as the price of its stated mission "to ensure the world safely makes the transition through transformative AI." Helen Toner of Georgetown's Center for Security and Emerging Technology likened the company's worldview to a high-stakes wager on powerful AI.

RESEARCH & MODELS

• MODELCNBC | OpenAI and Anthropic face new AI reality as companies shift from tokenmaxxing to efficiency | CNBC reported a growing crop of US founders is reining in AI spending after bills ballooned into the billions since ChatGPT's 2022 debut. AI startup Lindy CEO Flo Crivello switched 100% of his company's traffic off Anthropic's Claude to China's open-weight DeepSeek, saying the cost curve "crash[ed] to the ground" and would save millions within months. D.A. Davidson analyst Gil Luria warned some large enterprise customers "may start limiting their out-of-control token spend," signaling a shift away from the "tokenmaxxing" era of AI-assisted coding.
• MODELCaixin Global | Year in Review: DeepSeek's Breakout Rewrites U.S.-China AI Race | Caixin Global, in a year-end retrospective recirculated June 26, recounted how DeepSeek — backed by quant hedge fund High-Flyer Quant, one of only six Chinese firms with access to over 10,000 GPUs — emerged during China's Spring Festival as a major challenger. Its chatbot saw a download spike exceeding 2,000% to top global app-store rankings, and its open-source V3 and R1 models matched OpenAI's GPT-4 and o1 through algorithmic optimizations that lifted Nvidia GPU efficiency and lowered training costs, prompting rapid adoption by Tencent and Alibaba.
• SCIThe Wall Street Journal | InSilico Is on a Mission: Be No. 1 in China, Create 'God Drug' With AI | The Wall Street Journal reported AI drug-discovery firm InSilico Medicine, which has been developing treatments for cancer, Parkinson's and pulmonary fibrosis, is pursuing longevity therapies and a "China for China" strategy as it aims to become the top AI drug developer in the country.
• MODELVentureBeat | AI agent memory: MRAgent cuts token use up to 27x | VentureBeat reported researchers at the National University of Singapore developed MRAgent, an agent-memory framework that abandons the static "retrieve-then-reason" approach for a dynamic, multi-step memory reconstruction integrated into the model's reasoning. The technique cuts token consumption by up to 27x and lowers runtime costs versus other agentic memory-management approaches by letting an agent revise its retrieval strategy mid-reasoning rather than relying on fixed similarity scores.

COMPUTING & INFRASTRUCTURE

Still in play

• OpenAI's Jalapeño custom inference chip, built with Broadcom, with OpenAI joining Google, Apple and SpaceX in building custom silicon to hedge single-supplier reliance on Nvidia (TechCrunch).

MARKETS

25 Jun 2026 close | Retrieved 25 Jun 22:00 UTC | Yahoo Finance

AI Equities (1D) | Nvidia 195.74 USD -1.6% | Microsoft 352.83 USD -3.5% | Alphabet 343.71 USD -0.5% | Meta 543 USD -2.7% | Amazon 227.01 USD -3.1% | Palantir 107.27 USD -5.5%

Semiconductors (1D) | AMD 533 USD +2.5% | TSMC 434.99 USD -1.3% | Broadcom 378.91 USD -0.8% | ARM 347.71 USD -3.2% | Super Micro 31.68 USD -2.4%

AI Infrastructure (1D) | CoreWeave 98.76 USD -2.1%

Indices (1D) | NASDAQ 25,359 -0.5% | SOX 13,941 +3.6%

Coverage: 26 Jun 01:00 – 27 Jun 01:00 UTC