AI Daily Brief: 16 June 2026

16 June 2026

The biggest AI story overnight has a twist: Anthropic has now disabled its two most advanced models for everyone - not just foreign nationals as initially ordered. Meanwhile SpaceX keeps rewriting market records, a City AM analysis crystallises why the UK has a genuine window in the AI race, and Pokémon Go scan data turns up in military drone navigation systems.

Anthropic disabled Fable 5 and Mythos 5 for everyone, not just foreign nationals

Since we reported on the US government's order to cut off foreign nationals from Anthropic's Fable 5 and Mythos 5 models in yesterday's briefing, a significant development has emerged: Anthropic has disabled both models for all users everywhere, including US citizens, because it says it has no reliable way to verify user nationality at the model level.

The US Commerce Department issued an export control directive citing national security concerns over a jailbreak vulnerability. Anthropic's response was blunt: since it cannot determine whether any given user is a US citizen or a foreign national, it pulled both models entirely rather than risk violating the directive. AWS confirmed Anthropic asked it to revoke access "for all users in all regions."

Anthropic has pushed back on the characterisation of the jailbreak, stating the vulnerabilities are "relatively simple" and that "other publicly-available models are able to discover them" without requiring any bypass. No universal jailbreak has been demonstrated. The company acknowledged "perfect jailbreak resistance is not currently possible for any model provider." Other Claude models remain available.

Our take: This is the AI industry's first experience of a full export control shutdown - not a feature restriction, but a complete model withdrawal triggered by a government directive. The implication for any business running critical workflows on Fable 5 or Mythos 5 is stark: frontier AI infrastructure is now subject to the same geopolitical risk as semiconductors or satellite technology. UK firms using Anthropic's most capable models via AWS should be reviewing their resilience plans today.

SpaceX surges 20% on day two, pushing Musk's net worth past $1 trillion

SpaceX shares climbed another 20% on their second full day of trading on Monday, closing at $192.46 - more than 42% above the $135 IPO price. The move pushed the company's market capitalisation past $2.5 trillion, putting it within $135 billion of overtaking Amazon, and cementing Elon Musk as the world's first trillionaire with a net worth more than three times that of Google co-founder Larry Page.

SpaceX exercised its IPO over-allotment option, increasing total funds raised to $86.2 billion - the largest IPO in history. Retail investors bought as much SpaceX stock in the company's first two days of trading as they purchased across the entire US market in the preceding week, according to Vanda Research. The debut is widely read as a confidence signal for the planned Anthropic and OpenAI IPOs, potentially later this year.

The AI angle is direct: SpaceX's valuation is built substantially on Starlink's satellite internet business, which underpins AI infrastructure globally, and on growing US government AI and space contracts. The company's success has emboldened investors to push the broader AI rally further, with the tech-heavy Nasdaq 100 rising more than 3% on Monday following a US-Iran deal to reopen the Strait of Hormuz.

Our take: SpaceX's debut matters for AI for two reasons. First, it validates that markets will absorb massive AI-adjacent listings, clearing the path for Anthropic and OpenAI to go public at valuations that would have seemed absurd two years ago. Second, it concentrates an extraordinary amount of critical infrastructure - satellite internet, rocket launches, defence AI contracts - under a single founder whose political positioning is already influencing US AI export policy. UK businesses dependent on Starlink connectivity or AWS capacity should note that infrastructure risk and geopolitical risk are now the same risk.

The EU has regulated itself out of the AI race - and the UK still has a window

A sharp analysis piece in City AM today makes the case that the EU AI Act, celebrated in Brussels when it passed in late 2023, has delivered exactly what critics feared: a framework that prioritised regulating existing risks while the US and China raced ahead on compute, infrastructure and frontier models. A team of European researchers published a scenario paper called Europe 2031 last week, imagining a world where the US cuts Europe off from US-grown AI by executive order in 2029. Within days of that paper being published, the Trump administration did something very close to it, issuing the export control directive that forced Anthropic to withdraw Fable 5 and Mythos 5.

City AM argues the UK, freed from the EU AI Act by Brexit, is still in the game - but faces its own catching-up to do on compute and energy supply. The UK's £1.1 billion AI Hardware Plan and Sovereign AI Unit investments announced at London Tech Week last week represent a start, but analysts note the EU's trajectory is a cautionary tale about what happens when regulatory instinct consistently outpaces innovation instinct.

AI Summit London's 10th anniversary at Tobacco Dock drew 5,000 delegates and 300 speakers last week, with Minister for AI Kanishka Narayan framing the UK's challenge as ensuring AI benefits are broadly distributed while remaining globally competitive. London Business School Professor Michael Jacobides told the Summit: "AI is a strategy and organisation problem, not a technology problem."

Our take: The EU vs UK framing is useful but the real risk for UK businesses is complacency. Regulatory flexibility is only an advantage if businesses use the space to build capability rather than wait for certainty. The London Tech Week commitments look substantial on paper, but sovereign AI ambitions require sustained capital, talent and political will across multiple parliaments. The window is real. The question is whether UK businesses are moving fast enough to use it.

Pokémon Go scan data linked to military drone navigation AI

Niantic Spatial - the company spun out from the Pokémon Go creator Niantic - has been quietly building a visual positioning system using millions of 3D environment scans contributed by Pokémon Go players over the past decade. The system allows robots and drones to navigate using visual landmarks rather than GPS, making them effective even when GPS signals are jammed or spoofed. Reports published this week revealed the technology has been linked to military drone contracts.

The Decoder noted there is no public evidence that specific Pokémon Go scans will be used in any particular contract, and that GPS-jamming vulnerabilities are already well documented from conflicts in Ukraine and Iran. Niantic Spatial has not confirmed specific military applications. What is confirmed is that the database of spatial scans exists, it was built from consumer gameplay data, and the positioning system derived from it has attracted defence interest.

The episode raises a recurring question about consumer AI data: when users contribute data to a free product, they rarely understand the downstream uses that data might enable. Pokémon Go's terms of service do not explicitly prohibit military use of scan data.

Our take: UK businesses collecting spatial, behavioural or biometric data through consumer products should treat this story as a signal on data governance. The ICO's guidance on purpose limitation under UK GDPR is clear: data collected for one purpose cannot simply be repurposed for another, especially where the downstream use is materially different from what users would reasonably expect. If you are building any product that collects environmental scan data - from retail foot-traffic mapping to AR applications - your legal team needs to be thinking about this now, not when a contract surfaces.

Quick Hits

• At least 75 US AI data centre projects worth $130 billion were blocked or delayed in Q1 2026, according to a Data Center Watch report cited by NBC News.
• TSMC's CEO said chip supply is likely to remain tight for years because AI demand is still outpacing manufacturing capacity, supporting pricing power through at least 2027.
• AI Summit London's 10th anniversary drew 5,000 delegates and 300 speakers at Tobacco Dock, with the Minister for AI saying AI is "a strategy and organisation problem, not a technology problem."

Frequently Asked Questions

How often is the AI Daily Brief published?

Every morning at 7:30am UK time, covering the previous 24 hours of AI news from over 30 sources.

How are stories selected?

UK-relevant stories are prioritised first, then by business impact and practical implications for UK organisations adopting AI.

Why should business leaders follow AI news?

AI is moving faster than any technology in history. Staying informed is essential for making smart decisions about AI investment, adoption, and governance.