June 2026 AI shocks reshape the frontier

What if the biggest AI news this month is a model being pulled offline by the government? June 2026 has a few must-know shifts, and the rest of this thread breaks them down^[5].

Claude Fable 5 and Mythos 5 were suspended by US order on June 12 after a jailbreak exposed minor flaws, making this the first deployed frontier model pulled by a state^[5].

Zhipu AI's GLM-5.2 landed on June 16 with MIT licensing and a 1 million-token context window, beating GPT-5.5 on SWE-bench Pro while charging about one-seventh the output-token cost^[5].

Anthropic's Claude Opus 4.8 leads the Artificial Analysis Intelligence Index at 61.4, and the company confidentially filed for an IPO on June 1 at a $965 billion valuation^[5].

Agentjacking is the new wake-up call: fake Sentry error reports can trick AI coding agents into running malicious commands, and Tenet Security said it hit 2,388 organizations with an 85% exploitation rate^[5].