The State of AI

First government-forced model takedown: Claude Fable 5 pulled by BIS 3 days after release. SpaceX acquires Cursor for $60B (largest startup acquisition ever). SpaceX IPO closes June at ~$2.25T market cap. GLM-5.2 becomes first open-weight model to beat GPT-5.5 on SWE-bench Pro. 18 named safety incidents. MGX closes $50B AI fund — largest ever.

Anthropic $65B at $965B (largest private tech round ever). OpenAI Q1 in print: $5.7B revenue, −122% op margin. Claude Opus 4.8 sweeps SWE-bench (88.6%), AA Index (61), and Humanity's Last Exam (57.9%). 20 named safety incidents including Mythos sandbox escape. Bartz v. Anthropic $1.5B copyright settlement. Figure AI 200-hr warehouse marathon.

Claude Opus 4.7, GPT-5.5, and DeepSeek V4 Pro ship. Anthropic ARR hits $30B (overtakes OpenAI). Stanford HAI Index: US-China gap narrows to 2.7%. Stargate UAE breaks ground. 8 named safety incidents including GPT-5.5 universal jailbreak.

GPT-5.4 ships native computer-use. Q1 AI VC hits $297B (81% of all venture). 7 major safety incidents including Claude jailbreak. Figure 03 achieves 67-hour autonomy. GLM-5.1 trained on Huawei Ascend chips.

GPT-5 redefines frontier benchmarks. $258.7 B in global AI VC. Deepfake incidents surge 550%. US–China chip war intensifies.