China weighs AI export curbs as Hormuz attacks rattle peace talks

Two stories are dominating today's pulse: Beijing is considering restricting foreign access to China's most powerful AI models, and fresh attacks on shipping in the Strait of Hormuz are threatening fragile U.S.-Iran peace negotiations.

Chinese authorities have held talks with Alibaba, ByteDance, and Z.ai about limiting foreign access to their most advanced models, including unreleased systems, according to a Reuters report cited by TIME ^[1]. Officials have reportedly sketched options ranging from a ban on public release to domestic-use-only restrictions. That would mark a sharp reversal for Chinese labs, which have gained global traction by releasing open-weight models for free ^[1]. The shift appears driven by national-security concerns, mirroring recent U.S. moves to restrict access to Anthropic's Mythos and OpenAI's GPT-5.6 over their ability to find software vulnerabilities ^[1]. Tensions have been escalating: Alibaba recently banned Claude Code internally after a developer discovered Anthropic had embedded code to detect whether users were in China or linked to Chinese AI labs, a measure Anthropic said was aimed at combating model "distillation" ^[1].

Meanwhile, in the Middle East, a tanker caught fire off the coast of Oman after being struck by an "unknown projectile" in the Strait of Hormuz, the United Kingdom Maritime Trade Operations reported ^[2]. Citing unnamed U.S. officials, Axios reported that Iran's Islamic Revolutionary Guard Corps fired at least two missiles at commercial ships, damaging a Qatari LNG tanker and a Saudi-flagged crude oil tanker, with no casualties reported ^[2]. The attacks come during funeral processions for former Supreme Leader Ayatollah Ali Khamenei and as indirect U.S.-Iran talks in Qatar ended last week without clear progress toward a permanent deal ^[2]. The Strait of Hormuz, through which roughly 20 percent of global oil and gas supplies flowed before the war, has become a central bargaining chip, with Iran seeking service fees and the U.S. insisting on freedom of navigation ^[2].

In tech product news, Meta launched Muse Image, its first in-house AI image generator, on Tuesday ^[3]. Built by Meta Superintelligence Labs, the model powers image creation in Meta AI, Instagram Stories, and WhatsApp, with Facebook and Messenger support planned later this year ^[3]. Advertisers will soon be able to use it through Meta's Advantage+ platform ^[3]. The launch is part of Meta's broader push to bring AI capabilities in-house after losing some contracted Google Gemini capacity earlier this year ^[3].

Together, these stories underscore how AI and energy chokepoints are increasingly entangled with geopolitics.