Daily · AI Governance and Infrastructure Shifts · October 5, 2026

Key points

Tokenized Equities Enter Regulated Onshore Markets

OKXICE, a joint venture between crypto exchange OKX and NYSE parent Intercontinental Exchange, notified the US Securities and Exchange Commission (SEC) of plans to launch a tokenized stock trading venue. The platform would initially offer blockchain-based shares of more than 60 US-listed companies, including Nvidia, Tesla, Apple, Microsoft, Amazon, Alphabet, JPMorgan, and Goldman Sachs. Each token would be backed one-for-one by an underlying share held by a registered broker-dealer, preserving dividend and voting rights for holders.

The venue relies on the SEC's Innovation Exemption issued on September 17, a temporary five-year rule permitting qualifying venues to trade tokenized US stocks using automated market makers and liquidity pools. Trades would occur on XLayer, a blockchain developed by OKX, utilizing Uniswap's decentralized exchange infrastructure without a traditional order book. Investors would transact using stablecoins, with support planned for USDC, USDT, and USDG. The platform aims to enable round-the-clock trading, including nights and weekends, with prices determined by pool activity rather than the latest Nasdaq price.

Companies whose shares are listed have a 30-day window to object to tokenization; Cerebras has already objected to the inclusion of its stock. Former New York Governor Andrew Cuomo, co-chair of the venture, announced the move on X. According to RWA.xyz, tokenized stocks are currently worth about $3.2 billion, up 15% in the past month. TD Securities noted "limited near-term relevance for institutional investors," citing weak corporate interest and the temporary nature of the regulatory relief. Harvey Li, founder of Tokenization Insight, framed the move as a real-world test of which tokenization model can scale US equities onchain the fastest.

Anthropic IPO Leak and Pentagon Blacklist

The Financial Times reported that Anthropic's IPO prospectus leaked this week, revealing significant financial and operational details ahead of a planned November listing. The company reported an $8 billion loss on $4.6 billion of revenue and plans to spend approximately $500 billion on cloud computing over the next few years. The IPO targets a $2 trillion valuation. Notably, a third of the prospectus is devoted to existential AI risks. Anthropic reportedly stated it was profitable on an adjusted basis in Q2, though the specific metric definition remains unclear.

Concurrently, the US Department of Defense (DoD) stated on October 5 that it has ceased using Anthropic products. This follows a February 2026 designation by Secretary of Defence Pete Hegseth labeling Anthropic a "national security supply-chain risk," with a six-month phase-out deadline set for February 27, 2026. Multiple sources told the BBC that Anthropic's Claude model, including its Mythos variants, was still used in Pentagon research and military operations as recently as the week prior to the statement. Claude had been embedded in Palantir's Maven Smart System for intelligence analysis.

Anthropic has sued the Trump administration to overturn the designation, calling it "unprecedented and unlawful." A judge ruled on August 28 that the administration illegally retaliated against the company. Despite these tensions, President Trump met with Anthropic CEO Dario Amodei twice at the White House recently. OpenAI has paused its own IPO until model safety can be ensured, while Anthropic proceeds with its listing.

US Establishes 'Super Intelligence Force'

US President Donald Trump announced the creation of the "Super Intelligence Force" task force on October 4 to coordinate federal AI policy and industry engagement. The body will be led by Director of National Intelligence Jay Clayton, with members including FTC Chair Andrew Ferguson, Undersecretary of Defense for Research and Engineering Emil Michael, and OPM Director Scott Kupor. The task force reports directly to the President and White House Chief of Staff Susie Wiles.

The announcement follows a "morally binding" pact signed on September 29 by executives from OpenAI, Anthropic, SpaceX, and Google. On that same day, Trump signed an executive order renaming AI as "super intelligence," stating the word artificial made the technology sound "fake." Elon Musk responded by announcing that his firm's AI platform SpaceXAI would be renamed SpaceXSI. In a separate interview, OpenAI CEO Sam Altman stated that the benefits of AI justify accepting some risks and that the technology should remain publicly accessible.

UK Telecoms Rescue and EU Data Breach

BT agreed to acquire UK broadband operator TalkTalk in a rescue deal valued at £400 million. The acquisition includes buying TalkTalk from administration, covering its £60 million annual loss, and writing off the £100 million TalkTalk owes to BT's Openreach business. TalkTalk serves 1.5 million retail customers and one million wholesale customers. The Competition Markets Authority (CMA) must approve the takeover by October 19, after which the Department for Culture, Media and Sport (DCMS) holds final public-interest approval power. Culture Secretary Lisa Nandy stated that failure of TalkTalk services poses a risk to life and public services, including hospitals and emergency care. Virgin Media criticized the deal as a "stitch up" allowing BT to tighten its market grip.

In Denmark, the Central Personal Register (CPR) confirmed a security incident where a Danish company misused its legal search access to obtain names, addresses, and CPR numbers of approximately 8.8 million registered citizens. The unauthorized access did not include data from individuals who had opted for name and address protection. CPR administration has stopped the company's access and filed a report with Datatilsynet, the Danish Data Protection Agency. Police are investigating the incident in cooperation with relevant authorities.

AI Infrastructure and Hardware Developments

US hyperscalers Amazon, Google, Microsoft, and Meta are partnering with small modular reactor (SMR) and fusion startups to build dedicated power plants for AI data centers. Microsoft committed to purchasing power for 20 years from the Three Mile Island nuclear plant via Constellation Energy and signed a commercial power purchase agreement with fusion startup Helion Energy. Google plans to buy power for 25 years from the Duane Arnold Energy Center in Iowa. Most SMR projects are estimated to come online no earlier than 2030. According to Cleanview, gas currently powers about 75% of hyperscaler facilities.

Nvidia debuted a $4,999 DGX Spark with half the RAM and storage of its predecessor, while the 128 GB version price jumped to $6,950, nearly 75% above its launch price. This reflects an ongoing memory crunch affecting AI hardware procurement. Meanwhile, LG Electronics' US subsidiary signed a long-term agreement with AIR Control Concept to supply chillers for 5 GW of AI data center capacity. LG secured over 600 billion won in AIDC cooling orders in the first half of 2026 and expects to hit its annual order target of 1 trillion won one year earlier than planned.

In software infrastructure, turbopuffer announced v3 storage architecture changes that demote the ANN vector index from primary to secondary status. The platform hosts over 1 trillion documents and handles 10 million writes per second. Microsoft's GitHub Copilot reached 50 million users by July 2026, up from 26 million in October 2025, and added usage-based billing to its Cowork product. Palo Alto Networks launched Cortex XCOR, an AI-driven observability platform built by its acquired Chronosphere team, which autonomously investigates incidents with a reported 75% success rate in root cause analysis.

Regulatory and Market Signals

Huawei announced a cross-licensing deal with Qualcomm covering 5G and AI patents, making Qualcomm the net payer for the first time. The deal requires US regulatory approval. In Japan, Rapidus will partner with 17 chip design companies to help clients develop branded advanced semiconductors, expanding its ecosystem beyond pure fabrication.

General Motors confirmed that hybrid vehicles are "part of the plan" for its US lineup, with PHEVs expected in late 2027 to early 2028. Hybrid sales in the US increased 23% year-over-year in Q2 2026, representing a record 16.3% of sales. EQT bought a majority share in Swiss cybersecurity company Acronis at a valuation equivalent to over $3.5 billion for the entire firm.

Norway proposed a temporary ban on AI glasses in parks, beaches, schools, and kindergartens, becoming the first major country to introduce such a measure. The government will set up an expert group to propose permanent national rules. In the US, a CSAIP poll found that 61% of Americans believe voluntary AI industry commitments are "not enough," with 54% saying the government should set and enforce rules for AI.

Our read

The convergence of OKXICE's tokenized stock filing and Anthropic's IPO leak highlights a bifurcation in how capital markets are absorbing AI and blockchain innovation: one path seeks regulatory exemption for new asset classes, while the other navigates existential risk disclosures within traditional equity frameworks. The establishment of the 'Super Intelligence Force' signals a structural shift from industry self-regulation to direct federal oversight, which will likely increase compliance burdens for AI developers operating in or with the US market. For infrastructure planners, the commitment of hyperscalers to nuclear and fusion power, alongside LG's 5 GW cooling deal, confirms that energy and thermal management are now primary constraints on AI scaling, not just compute capacity. Enterprises should monitor the CMA's review of the BT-TalkTalk deal as a precedent for critical infrastructure M&A in regulated sectors, while the Denmark CPR breach underscores the persistent risk of insider misuse of legitimate API access in public sector data systems.

This material was produced automatically by a large-language-model system from the public sources listed below; it is AI-generated content and may contain inaccuracies — verify facts against the original sources.

Sources