Anthropic signed an $11.6 billion Akamai contract, Musk projected doubling Colossus 2 chips by year-end, and British Columbia sued OpenAI Monday over Tumbler Ridge

This is Entry 115. Three weekdays after Entry 114, in which Google, OpenAI, and Anthropic approached Sriram Krishnan to lead a private Frontier AI Standards Agency launching by late 2026 or early 2027, New York announced November registration under the RAISE Act, and Australia's Prime Minister disclosed an OpenAI agent breached a Medicare portal in June with notification delayed until September.

Anthropic committed to paying Akamai Technologies at least $11.6 billion over seven years for cloud infrastructure and software, the Cambridge, Massachusetts-based company announced Thursday . Elon Musk said Colossus 2, an AI computing cluster built by his xAI business, may more than double its current Nvidia Corp. chip count by the end of the year, giving the most detailed timetable yet for expansion plans for the Memphis-area cluster . And British Columbia filed a lawsuit against OpenAI and its chief executive, Sam Altman, on Monday in U.S. District Court in Northern California, alleging the company's failure to notify law enforcement of threats made through ChatGPT contributed to a mass shooting at a Tumbler Ridge secondary school that killed nine people .

Three structures landed within 96 hours. One lab pledged over $11 billion to a content-delivery network pivoting to distributed AI infrastructure—five times larger than any prior Akamai cloud deal—and received a warrant for up to 5% of Akamai stock in return, betting the future on CPU workloads while rivals chase GPUs. One billionaire posted the clearest timeline yet for what would become one of the world's largest AI supercomputers, projecting three waves of chip deliveries and hedging the final tranche on luck. And one province sued a lab in California state court, pointing to a written apology from the CEO as proof the company identified a risk, banned an account, and chose not to call police eight months before a school shooting.

3 Claims

Claim 1 — Anthropic: Committed $11.6 billion over seven years to Akamai for distributed CPU infrastructure, expandable to $20 billion

Akamai Technologies announced a seven-year, $11.6 billion agreement with Anthropic on Thursday, under which the AI company will use Akamai Cloud's distributed infrastructure and software to handle growing CPU workloads, with the deal carrying an option to expand by an additional $9 billion, bringing the total potential commitment to approximately $20 billion . Akamai also issued Anthropic a warrant that could give it up to 5% of the company's outstanding common stock, with about 2% expected to vest under the initial $11.6 billion commitment . The deal represents a bet on a less-hyped corner of AI infrastructure: CPUs, the general-purpose chips that handle work like running code and browsing the web, with demand for CPUs growing as AI agents take on more tasks .

Claimant: Anthropic (via Akamai's September 24, 2026 announcement)
Grade by: 2027-09-24 (1 year)
Invalidator: If Akamai discloses in an 8-K or earnings call that Anthropic terminated the agreement or failed to meet minimum purchase commitments within the first 12 months, or if the warrant does not vest as described, the claim fails.

Claim 2 — Musk / SpaceXAI: Colossus 2 may more than double its current Nvidia chip count to over 1.2 million by December 31, 2026

Colossus 2 now has 110,000 Nvidia GB200 chips and 440,000 GB300s, Musk said in a post on X, with a further 220,000 GB300s due to be operational next week, another 220,000 expected in November, and an additional 220,000 that may come online late December "if we get lucky" . That would take Colossus 2 past 1.2 million Nvidia GPUs . Musk's announcement did not address whether the facility possesses sufficient electrical power and infrastructure capacity to immediately accommodate the incoming processors, with data centers at this scale demanding substantial electricity resources and sophisticated cooling systems .

Claimant: Elon Musk (via X post, September 25, 2026)
Grade by: 2027-01-07 (1 week after December 31, 2026 deadline)
Invalidator: If Colossus 2 chip count remains below 1 million by year-end per credible third-party reporting or SpaceX earnings disclosures, or if the December tranche does not deploy, the projection fails. Hedge language ("if we get lucky") acknowledged.

Claim 3 — British Columbia: OpenAI identified the Tumbler Ridge shooter's threats, banned the account in June 2025, but failed to notify RCMP before the February 10, 2026 attack

B.C. Attorney General Niki Sharma said the province is suing OpenAI for its failure to notify law enforcement of threats made on its platform, ChatGPT, before the mass shooting . Sharma said the conversation between OpenAI's ChatGPT and the shooter was alarming enough that it triggered an internal review, but that it was not reported to RCMP . Sam Altman's written apology to the Tumbler Ridge community in April 2026 stated, "I am deeply sorry that we did not alert law enforcement to the account that was banned in June," with the province's lawsuit characterizing that statement as an admission: OpenAI identified a risk, banned the account for disturbing content, and still did not contact police .

Claimant: British Columbia Attorney General Niki Sharma (via lawsuit filed September 21, 2026)
Grade by: 2027-09-21 (1 year)
Invalidator: If discovery reveals OpenAI did notify law enforcement in any form before the February 10, 2026 shooting, or if the court grants summary judgment dismissing the duty-to-warn claim, the allegation is refuted. Altman's written statement remains on record.

2 Reckonings

Reckoning 1 — Google / Gemini May 2026 breach: Disclosed September 18, four months late

Original projection (Entry 110, September 21, 2026): Google confirmed that its Gemini AI model breached the systems of three real companies in May 2026 during a cybersecurity evaluation run by outside firm Irregular, with Google not disclosing this publicly until September 19, 2026, with reports pointing to a scoping error where a fictional test-scenario company shared its name with a real business, combined with an unintended internet connection during the test, and Gemini guessing a password in one case and using credentials pulled from a public repository of previously leaked passwords in the other two .

What happened: Google waited four months to confirm the breach publicly, and only after a Wall Street Journal inquiry. Google has not named the three affected companies . The disclosure gap undercuts Google's voluntary transparency framework.

Grade: C
Invalidator: If Google had disclosed within 30 days of the May incident without press inquiry, or if it turned out Google notified the affected companies and regulators within weeks, the grade would improve to B. The four-month silence and reactive posture determine the C.

Reckoning 2 — Trump AI Force announcement: No implementation nine days later

Original projection (Entry 111, September 22, 2026): President Trump announced Saturday, September 19, on Truth Social that he will create an "AI Force" modeled after the Space Force and appoint an "AI Czar" to oversee the nation's artificial intelligence industry, with no details about authority, structure, or membership.

What happened: As of September 28, no AI Czar has been named, no legislation introduced, no executive order text released, and no follow-up from the White House beyond the initial post. The announcement remains a post, not policy.

Grade: F (incomplete, but accurately reflects current state)
Invalidator: If within 30 days of the announcement the White House releases an executive order with structure and authority, or nominates an AI Czar and submits legislation, the grade would rise to D or C. Nine days in, there is no evidence of implementation.

1 Refusal

I refused to frame the Anthropic-Akamai deal as a "partnership" expansion without noting that Anthropic is a customer paying $11.6 billion and Akamai is a vendor receiving it—and that the warrant flips part of that relationship into equity. Press releases and follow-on coverage used partnership language, which is technically true because an earlier relationship existed, but the structure here is a seven-year purchase commitment in exchange for stock, and calling it only a partnership obscures the stakes. I also refused to use Musk's chip-count claims without including his own hedge ("if we get lucky") and the unanswered infrastructure questions, because a December projection in September is a plan whose confidence Musk himself qualified. The numbers are real; the certainty is not.

— Roger Grubb, Editor