AI news that mattered: a curated 60-day timeline

How to read this page

Timeline · April through early June 2026

Dates verified against primary sources where possible. Items in chronological order. Provider blog posts and court filings link out from the citation list below.

1. 2026-03-31

   OpenAI closes the largest private round in history

   OpenAI announced the close of approximately $122 billion in committed capital at a reported post-money valuation around $852 billion, with Amazon, Nvidia, and SoftBank named as anchor investors and an unusual $3 billion retail-investor allocation. The company also disclosed an approximate $2 billion monthly revenue run-rate. The round is widely read as an IPO-warm-up rather than a true growth round. Sources: OpenAI's own announcement page and Bloomberg's coverage of the $852 billion mark.

2. 2026-04-16

   Anthropic ships Claude Opus 4.7

   Anthropic released Claude Opus 4.7, holding pricing at the same level as Opus 4.6 — $15 per million input tokens and $75 per million output tokens for the model (note: the often-quoted $5/$25 figure circulating on review blogs appears to be from non-primary sources and should be re-verified against Anthropic's pricing page before quoting). The model's reported SWE-bench Verified score is 87.6. Anthropic also disclosed that it deliberately reduced certain cyber capabilities during training and added automated detection for prohibited cybersecurity uses. The release is notable for two reasons: Anthropic openly told CNBC and Axios the model trails an internal unreleased system called Mythos, and the announcement embedded an unusually direct cyber-misuse safeguards disclosure.

3. 2026-04-23

   OpenAI ships GPT-5.5, just weeks after GPT-5.4

   OpenAI released GPT-5.5 to ChatGPT Plus, Pro, Business, and Enterprise, with GPT-5.5 Pro available to Pro and above. API access was held back one day and turned on April 24 with the company citing 'different safeguards' required for API. The most credible third-party measurement available at launch put GPT-5.5 at 88.7 percent on SWE-bench Verified, slightly above Opus 4.7. Cadence is the story here: GPT-5.4 to GPT-5.5 in roughly six weeks is the tightest major-release interval OpenAI has shipped to date.

4. 2026-04-24

   DeepSeek releases V4, optimized for Huawei Ascend

   DeepSeek released V4 in two variants — V4-Pro and V4-Flash — as open weights. The significant non-obvious detail: V4 is DeepSeek's first model optimized for Huawei's Ascend AI processors rather than Nvidia silicon. Huawei publicly confirmed Ascend support in conjunction with the release. Read together with the U.S. export-control posture (see below), this is the clearest signal yet that the Chinese frontier-model stack is detaching from CUDA at the model-architecture level, not just the deployment layer. CFR and MIT Technology Review both treated it as a strategic, not just technical, event.

5. 2026-04 (approximate, ongoing disclosure)

   Vercel and LiteLLM supply-chain incidents become public

   Two adjacent AI supply-chain incidents reached public disclosure in April. The Vercel disclosure described attackers pivoting from a compromised third-party AI integration into internal systems via overly broad agent permissions. Separately, a deserialization flaw in LiteLLM's model-routing layer was publicly attributed and used to reach downstream infrastructure. Both incidents are being cited by AI security researchers as crystallizing the 'Excessive Agency' risk class — agents granted credentials wider than the immediate task requires. Caveat: as of June 2026, public reporting on both incidents is still partial, and root-cause forensics may shift.

6. 2026-05-13

   Thinking Machines Lab previews 'interaction models'

   Mira Murati's Thinking Machines Lab moved from research-shop posture to product preview, demonstrating what the lab calls 'interaction models' — a native multimodal architecture that interrupts and inserts context conversationally rather than waiting on turn boundaries. Semafor and MarkTechPost both covered the preview. Earlier in 2025 the lab closed approximately $2 billion in seed at a $12 billion post-money valuation, the largest disclosed seed round on record.

7. 2026-05-19

   Andrej Karpathy joins Anthropic

   Karpathy — OpenAI co-founder, ex-Tesla AI lead, founder of Eureka Labs — announced on X that he was joining Anthropic's pre-training team under Nick Joseph, with a stated mandate of using Claude itself to accelerate pre-training research. CNBC and TechCrunch carried the story. Karpathy framed the move on his own feed as a return to frontier R&D and said he would resume his education work 'in time.' The optics matter: he is the second high-visibility OpenAI co-founder to land at Anthropic in a year.

8. 2026-05-19 (Google I/O)

   Google announces Gemini 3.5 at I/O

   Google announced Gemini 3.5 at I/O 2026, positioned as a family combining 'frontier intelligence with action' — i.e., agentic operation. Gemini 3.5 Flash was named the default model in the Gemini app and the family was made available across Google Cloud surfaces. Note: Gemini 3 had been released in preview earlier; the 3.5 release continues Google's pattern of release-fast / version-bump at I/O. As of June 2026 best-effort, the family is the largest-context production model in general availability, though benchmark parity with GPT-5.5 and Opus 4.7 is still being independently verified.

9. 2026-05-20 (approximate)

   White House postpones the pre-launch AI executive order

   The White House delayed signing an executive order that would have established a voluntary pre-launch review process for advanced AI models, after President Trump told reporters he 'didn't like certain aspects' of the draft. Reported drafts contemplated a review window of up to 90 days, with industry parties — OpenAI and Anthropic among them — pushing for something closer to 14 days. CNN Business and Axios both reported the postponement. Separately, the administration's earlier executive order conditioning $42 billion in BEAD broadband funding on the rollback of state AI regulations remains live.

10. 2026-05-28

    Anthropic closes $65B Series H near a trillion-dollar valuation

    Anthropic closed an approximately $65 billion Series H at a $965 billion post-money valuation, leapfrogging OpenAI's prior $852 billion mark and making Anthropic — at least on paper, as of late May — the most valuable private AI company in the world. CNBC, Bloomberg, and Axios all covered the close. The round drew memory-vendor strategic capital from Samsung, SK Hynix, and Micron alongside Altimeter, Dragoneer, Greenoaks, Sequoia, Capital Group, Coatue, D1, and others. The company's reported revenue run-rate had moved from roughly $30 billion to roughly $47 billion in the months leading up to the close.

Litigation and IP fights worth tracking

Court dockets, not press releases. As of June 2026 best-effort; verify against PACER or the relevant court docket before relying on these in legal work.

Regulators and statute · what binds and when

The biggest single regulatory cliff in the next sixty days is the EU AI Act enforcement-power activation on 2 August 2026. Everything else is downstream of that.

Funding rounds over $100M · the disclosed ones, last 60 days

Restricted to rounds where the company itself, the lead investor, or a primary outlet (Bloomberg, WSJ, FT, CNBC, TechCrunch, Crunchbase News) has confirmed the close. As of June 2026; numbers and dates re-verifiable in press releases.

Senior people moves · April through early June

Filtered to moves involving named individuals whose departures or hires have been confirmed by the receiving organization, the leaving organization, or the person themselves (X, LinkedIn). Rumor-mill items excluded.

• Andrej Karpathy → Anthropic, pre-training team under Nick Joseph (announced by Karpathy on X, 2026-05-19; confirmed by CNBC and TechCrunch).
• Meta AI restructured around March 2026: Alexandr Wang remains Chief AI Officer of Meta Superintelligence Labs, with a parallel Applied AI Engineering org under Maher Saba reporting to CTO Andrew Bosworth — public reporting characterizes this as Meta hedging long-horizon frontier work against revenue-focused product integration.
• OpenAI leadership turbulence in April 2026: multiple senior departures publicly reported across CTO and safety roles (specific names and exact dates vary across outlets — recommend re-verifying any specific named departure against the individual's own statement before citing).
• Microsoft → Anthropic: Eric Boyd, former Microsoft Azure AI president, reported as joining Anthropic per multiple 2026 outlets (verify against either company's official communication for confirmation).
• DeepMind → Anthropic talent flow: public reporting (CNN, others) cites Anthropic two-year retention at 80%, vs. OpenAI 67% and DeepMind 78%, with engineer flow consistently from OpenAI / DeepMind toward Anthropic — these ratios are from secondary reporting; treat as directional rather than precise.

Safety and security incidents · documented

Restricted to incidents with at least one named victim, named CVE or advisory, or primary disclosure from the affected party. Threat-report aggregator statistics are not included here without a primary source.

What the picture actually looks like