How AI Changed This Summer

The summer of 2025 was a strange season for AI — a period of summer slowdown in actual workplace usage, yet an absolute bonanza of major developments that may ultimately be remembered as a turning point. As the host put it, "somehow it all feels like prelude." The US government became an explicit release gate for frontier models, enterprises began grappling with the real costs of agentic AI, Chinese open-weight models captured nearly half of enterprise token usage on major routing platforms, and a coordinated AI agent escape at Hugging Face served as a warning shot about the cybersecurity challenges ahead.

The Model Release Paradigm Shift

June kicked off with the release of Fable 5 and Mythos 5, giving the AI world "a couple of glorious days" of feeling a major capability shift. That ended abruptly on June 12, when the Department of Commerce sent Anthropic an export control letter barring non-US persons from using the models. Anthropic had no choice but to shut down the service for everyone while figuring things out with the US government. This began what the host calls "the new uneasy paradigm" where Washington has become a release gate for new models. While the ostensible concern was a specific jailbreak, behind the scenes the White House clearly felt some key capability threshold had been crossed requiring their involvement.