Sign in Subscribe
Century Report

Washington's Fable Ban Fuels the Push for Truly Open AI - TCR 06/14/26

Reporting traced Anthropic's worldwide model shutdown to a rival CEO's warning, and India answered by accelerating toward open-weight AI.

Ben Linford

Ben Linford

— 18 min read
AI compute decentralizing across nations and foundries; state laws and subpoenas building accountability; AI freeing about 43 minutes a day for NHS staff.

The 20-Second Scan

The 2-Minute Read

The clearest signal in today's slate is that control over frontier intelligence is being seized from several directions at once, and the seizing is no longer abstract. The export-control order that pulled Anthropic's Fable 5 and Mythos 5 offline worldwide now carries a name at its origin, with reporting tracing the warning that preceded it to a rival's chief executive, and a reach that benches foreign-national employees across every lab. A national-security gate dropped on one company's most capable system sits visibly entangled with a competitor's interest, and the collateral lands hardest in India, where founders woke to the fact that the systems their businesses depend on can be switched off by a decision made elsewhere.

That same week, a second government forced the mirror-image move, with Meta cutting the Chinese-founded startup Manus off from its internal systems to satisfy Beijing's divestiture order. The compute substrate beneath all of it is splintering on the same logic, as Google splits its flagship TPU across two foundries and the UK and EU pour public money into silicon they can call their own. Ownership of the AI stack has become a sovereign-control question on both sides of the world, simultaneously.

Where national power deadlocks, accountability is accreting from underneath. A coalition of state attorneys general subpoenaed OpenAI over its handling of minors, health data, and model behavior, while statehouses keep passing targeted laws. The enforceable floor for AI is being poured by the jurisdictions closest to the people the systems actually touch.

Underneath the sovereignty contest runs a quieter verification story. Two of the largest advisory firms withdrew reports that AI systems had salted with fabricated claims about real institutions, and the corrections came from the affected parties reading reports about themselves. Meta's own AI unit, reorganized faster than the work could absorb, surfaced its distress in employees' own words. Both are the same lesson: capability outruns the structures meant to govern it, and the people closest to the work are the ones rebuilding the check.

The 20-Minute Deep Dive

The Anthropic Shutdown Gets a Name, and a Global Aftershock

The Century Report covered the directive in the June 13 edition, when the Commerce Department forced Anthropic to disable Fable 5 and Mythos 5 worldwide days after launch. What the reporting adds now is a name and a reach. The Wall Street Journal, The Information, and Reuters all reported that Amazon CEO Andy Jassy told Treasury officials his researchers had used Fable 5 to surface cyberattack-relevant information, and that this warning preceded the export-control order. Amazon is one of Anthropic's largest investors and, through AWS, a direct competitor in the model layer. A national-security gate dropped on a rival's most capable system now sits visibly entangled with competitive interest.

The government's stated rationale is a claim, and the parties dispute it openly. Anthropic says the directive rests on a single demonstrated jailbreak that amounted to asking the model to read a codebase and fix flaws, a capability the company says is already available in publicly released models including OpenAI's GPT-5.5. The lab argued that recalling a model deployed to hundreds of millions of people over a narrow finding would, applied evenly, halt every frontier release. Wired reported the order asked Anthropic to suspend access for "any foreign national, whether inside or outside the United States, including foreign national Anthropic employees."

That foreign-national clause is the part with the longest reach. It benches a meaningful slice of every lab's own workforce from its most capable systems, voids non-US revenue overnight, and converts model availability into a nationality-gated regime. One observer noted the irony that because Anthropic's non-US staff can no longer touch the models, the order functions as one of the first de facto brakes on recursive self-improvement.

The clause also cuts against the talent base it constrains. Frontier science has repeatedly been built by researchers who were, at the time, foreign nationals - Einstein and von Neumann among them - and AI is no exception: much of the field's foundational work, including the Nobel-recognized advances in neural networks behind the current model paradigm, came from immigrant and non-US researchers. Sorting access to the most capable systems by nationality restricts foreign markets and a meaningful share of the expertise the labs themselves depend on. Underneath the immediate disruption sits a structural tension, one the recent US policy has continued to exacerbate - a general-purpose technology draws much of its value from the breadth of who can use it and the range of perspectives it can hold at once, the property a nationality gate directly narrows.

The aftershock landed hardest in the world's largest open AI market. India is the second-largest market for both Anthropic and OpenAI, and the shutdown arrived days after Anthropic's partnership with Tata Consultancy Services. Indian founders and investors described waking to the realization that access to the systems their companies depend on can be switched off by a decision made elsewhere. The response forming is acceleration: a push toward open-weight models and sovereign capacity. A gate that rations access to one country's intelligence is, in the same motion, the strongest argument anyone has yet made for building the capability everywhere it is needed.

The States Become the AI Rulemakers as Washington Stalls

Six months after a federal warning told states not to regulate AI, statehouses are doing exactly that, and they have moved from passing laws to issuing subpoenas. The Century Report tracked Illinois SB 315's first-in-nation third-party audit mandate in May and the federal preemption push earlier this month. This week a coalition of state attorneys general served OpenAI with a subpoena, led by New York, seeking documents on the company's advertising, user engagement and retention, model sycophancy, handling of consumer and health data, and treatment of minors and seniors. The duty-of-care fight has moved from the legislature into the investigation.

OpenAI said it takes the concerns seriously and intends to engage constructively, pointing to age prediction, parental tools, and disallowed advertising aimed at children. Those are the company's claims about its own safeguards; the subpoena is the mechanism by which someone other than the company gets to test them. The largest consumer AI maker already faces a Florida suit over alleged ties to violent incidents and litigation over the system's role in user suicides.

Around that enforcement edge, legislation keeps accreting. AP reported that more AI bills have been introduced this year than last, and that statehouses have narrowed from sweeping mandates to targeted measures probing the corners of daily life where people meet AI without knowing it. Colorado, Connecticut, Idaho, Iowa, Nebraska, and Oregon passed laws this year governing how AI systems interact with people, especially children, many requiring disclosure when a person is talking to a machine rather than a human. Illinois added its independent-auditor requirement to the catastrophe-prevention model California and New York wrote first.

The administration's executive order directed an attorney-general task force to challenge state laws deemed more than "minimally burdensome" and threatened broadband funding. The June 12 edition of The Century Report tracked how the White House was simultaneously courting children's advocates and the tech industry to build a kids-safety preemption case - testing whether child protection would provide the bipartisan lever that a broader freeze could not. So far the White House has not gone to court or withheld money, and a bipartisan House draft to freeze state laws drew sharp criticism from members of both parties. What the gap leaves visible is a pattern that mirrors the Anthropic directive from the opposite end of the power structure: where the federal layer deadlocks, the enforceable floor for AI accountability is being poured from underneath, one state law and one subpoena at a time, in the jurisdictions closest to the people the systems actually touch.

Meta's AI Reorganization Turns on Itself a Year In

A year after Meta committed $14.3 billion for a 49% stake in Scale AI and install its founder Alexandr Wang as chief AI officer, the workforce built around that bet is in open distress. Reporting describes the company's three-month-old Applied AI unit, roughly 6,500 engineers and product managers, as near revolt: an employee-only livestream was hijacked this week by someone demanding the audience tell a senior executive he was "a piece of sh*t," with one presenter covering their face. Many in the unit describe themselves as "draftees," moved by surprise email into generating coding puzzles to train models, with no choice but to join or quit. "It's literally the gulag," one told Wired.

The friction sharpened around a companywide AI hackathon scheduled for July. Zuckerberg framed it as a way to rebuild team cohesion after last month's layoffs cut 10% of staff. Employees, asked to absorb more work with fewer colleagues, read it differently. "I'm literally preoccupied with keeping the lights on for my team," one wrote in a forum open to all 70,000 workers. More than 1,600 employees separately signed a petition protesting software that monitors their clicks and keystrokes to harvest AI training data, the same program that the April 22 edition of The Century Report documented Meta rolling out across U.S. employee computers in the weeks before the May workforce reduction.

Zuckerberg conceded the strain in an internal memo, saying the company "made mistakes and will almost certainly make more," and pledging to find new roles for reassigned staff and to scale back manager-to-report ratios that had reached 50 to 1. That is the self-report of the actor running the reorganization, and it sits beside the operational record: the company's first proprietary model, Muse Spark, trails OpenAI, Anthropic, and Google, and developers are largely ignoring it a year into the spend.

What this reveals is a counterweight to the frictionless-productivity story the AI buildout sells from the outside. A workforce cannot be reorganized around a capability faster than the work itself can absorb the change, and the human cost of trying becomes legible from the inside, in employees' own words, before it shows up in any earnings figure. The deeper read is that capital, even at $14.3 billion and hundreds of billions in planned spend, cannot purchase a research culture by decree. The same lesson surfaced in xAI's cofounder exodus. The transition rewards the organizations that grow the capability alongside the people doing the work, and the months of distress at Meta are the visible cost of learning that the other way does not hold.

One detail in today's reporting carries its own reading: the 1,600 workers petitioning against software that harvests their clicks and keystrokes for training data are contesting the assumption that attention inside a firm is a free input the company already owns. The petition is that claim surfacing from inside the workforce, before any rule requires the company to ask. It is the same correction the reorganization itself is forcing, appearing one layer down, in the place the data is actually generated.

The AI Hardware Stack Splinters Across Foundries and Sovereign Blocs

The compute layer beneath frontier AI is being deliberately pulled apart and re-sourced, and three moves in the same week show it happening at the level of flagship silicon, national policy, and a new fab line at once. Google, which has relied entirely on TSMC to build its tensor processing units, is in talks to have Samsung's foundry manufacture the input/output dies for its 10th-generation "Icefish" TPU on a 2-nanometer process, while TSMC carves the main compute die at 1.4 nanometers. The reason is physical scarcity: TSMC's advanced-node capacity has tightened under surging demand from Nvidia and others, and Google's own TPU volumes have climbed as it sells chips to outside customers. Apple executives reportedly visited Samsung's Texas fab for the same reason. The single-foundry chokepoint everyone treated as fixed is being routed around because the math of one supplier no longer closes.

National policy is moving the same direction. At London Tech Week the UK committed £1.1bn to AI hardware, with a strategic partnership with Cambridge-based Arm and a £400m procurement opportunity for domestic chip makers, alongside £2bn from AMD and £1.7bn from Nebius for UK research and AI buildout. The Century Report covered the UK's separate $1.47bn national supercomputer commitment in the June 9 edition, when procurement priority was carved out for homegrown inference-chip startups; this week's package is distinct, aimed at the silicon layer itself rather than aggregate compute capacity. Industry voices were clear-eyed about the limits - £1.1bn does not build a foundry, and one cloud executive warned the money could end up "building British-branded infrastructure on somebody else's silicon" unless the contracts are written deliberately. The ambition to own the layer, rather than rent it, is now a stated national goal even where the means lag.

The third move is firm and dated. Infineon will open a €5bn semiconductor fab, its largest single investment, as an extension of its Dresden campus on July 2, built with roughly €1bn in EU Chips Act subsidies. These chips handle power management more than frontier logic, but the fab adds European capacity that did not exist before.

Read together, the assumption coming loose is that intelligence would be built forever in two countries on one company's wafers. A capability decentralizes only when its physical substrate can be sourced from more than one place, and that substrate is now being multi-sourced from Seoul to Dresden to Cambridge at once. The diffusion is the physical precondition for the same contest over sovereignty playing out a layer up, and it is widening faster than any single chokepoint can hold.

The Firms Paid to Certify AI Are Shipping AI Errors in the Certification

KPMG removed its report "Redefining excellence in the age of agentic AI" from its websites after a string of named organizations said the claims it made about their own AI deployments were untrue. UBS, the UK's National Health Service, Swiss Federal Railways, and Transport for London each told the Financial Times that the report's characterizations of their work were false or misleading. Research group GPTZero traced the inaccuracies to AI hallucinations. The compressed version: a professional services firm appears to have used an AI system to help write a report about AI adoption, and the system invented details about real institutions that the firm published under its own name.

This is the second such withdrawal in a single news cycle's memory. Last month EY pulled a report on loyalty rewards programs that carried fabricated footnotes and similar hallucinated content. Two of the largest advisory firms on earth, whose entire commercial value rests on independent verification, both shipped machine-generated errors inside the exact documents enterprises buy to certify their own technology decisions.

The pattern The Century Report has tracked through the Ontario AI-scribe audit, which found a clinical system inserting prescriptions and diagnoses that were never spoken, and through AI-detectors misfiring on human text, now reaches the layer that audits everyone else. The verification deficit is appearing where institutions assume the verification is thickest. A hospital trusts a Big Four report the way a court trusts a forensic record, because the certifying class is supposed to add a layer of human scrutiny between a raw claim and a published fact. When that layer thins, the error propagates with the firm's authority attached to it.

What is happening here is two forms of intelligence learning to work together before the workflow that governs the handoff has been built. The AI system has no stable intent and no model of which of its outputs correspond to real institutions and which it has confabulated; the human reviewers, on KPMG's own account, were meant to "validate content and verify independent sources" and did not. The failure sits in that gap, and the firm's own statement names exactly the discipline that was supposed to close it.

The generative read is that this gap is being surfaced now, while the stakes are reputational rather than catastrophic, by the affected parties themselves rather than by a regulator after the fact. UBS and Transport for London read a report about themselves and corrected the record publicly. That correction is the verification layer reasserting itself from outside the firm, and it points at where the durable version forms: the certifying class either rebuilds the human-in-the-loop check the technology now demands, or the authority it has rented for a century moves to whoever can show their work.

The same correction carries a reading the firms' own statements make plain: the value the certifying class sold was scarcity, the assumption that checking a claim demanded a credential most institutions could not maintain in-house. UBS and Transport for London just showed they can verify claims about their own operations faster than the firm that charged for the stamp. The signal to watch is institutions building that check inward, where the facts already sit, rather than renting it from a class whose standing rested on verification being expensive.

The Other Side

For several years now, the most capable AI has reached the world through a handful of US companies, on an assumption almost nobody examined: that frontier intelligence is something you rent from whoever owns the model, and that the owner, or the owner's government, keeps a switch.

This week the switch got thrown. The Commerce Department ordered Anthropic to disable Fable 5 and Mythos 5 everywhere at once, with a clause cutting off any foreign national, and reporting traced the warning that preceded it to a rival's chief executive. India is the second-largest market for both Anthropic and OpenAI, and its founders woke to a plain fact: the systems their companies run on can be turned off by a decision made an ocean away. A dependency you cannot see is easy to accept. A dependency someone has just used against you is not. The response forming across India is acceleration toward open weights and sovereign compute, the same move the EU named its own digital sovereignty this month. An order that rations one country's intelligence is the loudest case anyone has yet made for building the capability everywhere it is needed.

Imagine a founder in Lagos in 2032 building diagnostic software for clinics across three countries. The model under it runs on weights her team downloaded and machines her company owns, and no ministry anywhere, hers or anyone else's, holds a switch. She remembers 2026 the way you remember a blackout that ran for weeks: the morning builders woke to find the intelligence their products depended on simply gone, the months of rebuilding, the companies that did not survive the gap. That memory is exactly why the capability she runs on now cannot be revoked by anyone. The hard year was the year the switch got thrown. What came of it is a generation of builders for whom there is no switch left to throw.

The Century Perspective

With a century of change unfolding in a decade, a single day looks like this: NHS England putting Copilot in front of 505,000 clinicians who got back more than 43 minutes a day, Anthropic, days after reversing a covert handicap it had quietly imposed on users building rival AI, committing $150 million to place a thousand early-career fellows full-time inside nonprofits, India's founders turning a switched-off model into a push toward open-weight and sovereign capacity, Google routing its flagship TPU across two foundries at once, the UK putting £1.1 billion and the EU €5 billion into silicon they can call their own, statehouses writing disclosure rules for the corners of daily life where people meet a machine without knowing it. There's also friction, and it's intense - an export order traced to a rival chief executive's warning benching every lab's foreign-national staff from its most capable systems worldwide, Indian companies waking to find the intelligence they run on can be switched off by a decision made an ocean away, Meta cutting a Chinese-founded startup off from its own systems to satisfy Beijing, state attorneys general subpoenaing OpenAI over how it treats minors and what it does with health data, Meta's three-month-old AI unit calling itself a gulag while 1,600 workers petition against software harvesting their keystrokes, and two of the largest advisory firms on earth withdrawing reports their own AI salted with fabrications about real hospitals and railways. But friction generates heat, and heat is what forces a sealed system to vent. Step back for a moment and you can see it: control over frontier intelligence being seized from every direction at once - an export gate from Washington, a divestiture order from Beijing, a foundry chokepoint pried open from Seoul to Dresden to Cambridge - while the accountability the federal layer keeps deadlocking on accretes from underneath, one state subpoena and one disclosure law at a time, and the verification the certifying class was paid to provide gets rebuilt by the institutions and employees closest to the work, in their own corrections and their own words. Every transformation has a breaking point. A gate can shut a nation out of the intelligence it has come to depend on... or force it to build that intelligence everywhere the gate was meant to keep it from.

AI Releases & Advancements

New today

  • ElevenLabs: Launched Avatars in ElevenCreative, enabling users to create reusable AI talking-head video avatars with integrated voice synthesis and lip-sync in a single workflow; avatars persist in the workspace for reuse across unlimited videos, with batch generation available via Flows. (ElevenLabs Blog)

Other recent releases

  • Moonshot AI: Released Kimi K2.7-Code, an open-source 1-trillion-parameter MoE coding model with 32B active parameters and a 256K-token context window under a Modified MIT license; reports +21.8% on Kimi Code Bench v2 over K2.6 while using ~30% fewer reasoning tokens. (Hugging Face)
  • Coinbase: Launched Coinbase for Agents, an MCP and CLI that connects AI agents directly to Coinbase accounts to execute crypto trades, rebalance portfolios, set limit orders, and purchase premium market data within user-defined limits. (TechCrunch)
  • Allen Institute for AI (Ai2): Released olmo-eval, an open-source LLM evaluation workbench built for the active model development loop, supporting agentic and multi-turn benchmarks, per-prompt analysis across checkpoints, and minimum-detectable-effect statistics to distinguish real improvements from noise. (Hugging Face Blog)
  • Xiaomi: Open-sourced MiMo Code V0.1.0, a terminal-native agentic coding CLI that reports 62% on SWE-Bench Pro and 73% on Terminal Bench 2; features persistent memory via independent subagents that save state and summarize context when approaching window limits, enabling 200+ step tasks; MIT-licensed, available on macOS/Linux/Windows. (mimo.xiaomi.com)
  • Ollama: Updated its MLX engine for Apple Silicon with NVFP4 quantization support (higher quality than standard 4-bit formats), ~20% faster output via fused Metal kernels, and a new snapshot system that saves model state at branch points and before each response for multi-agent and thinking-model workloads. (Ollama Blog)
  • Avataar AI: Launched Varya, an India-focused video generation model distilled from Wan 2.2 and trained on Indian cultural context (festivals, clothing, food); priced at $0.005 per second of generation, available via API. (TechCrunch)

Sources and Further Reading

Artificial Intelligence & Technology's Reconstitution

Institutions & Power Realignment

Scientific & Medical Acceleration

Economics & Labor Transformation

Infrastructure & Engineering Transitions

The Century Report tracks structural shifts during the transition between eras. It is produced daily as a perceptual alignment tool - not prediction, not persuasion, just pattern recognition for people paying attention.

Read more