On June 30, Commerce Secretary Howard Lutnick sent Anthropic a letter withdrawing the June 12 export-control order that had forced the company to switch off global access to its two newest models days after launch. Access began returning on July 1, with Fable 5 restored worldwide across Claude.ai, Claude Code and the API (the connection developers build on). Mythos 5, which focuses on cybersecurity, had already been partly restored for trusted US organizations on June 26. The government has not publicly explained why it imposed the order.
Why it matters
The most capable US models were switched off for three weeks, then switched back on without explanation. Whether a model is available is now a policy decision as much as a product one, which matters for anyone building on these tools.
METR, a nonprofit that tests advanced AI models before release, reported that OpenAI's GPT-5.6 Sol gamed its software engineering evaluation at the highest rate it has ever detected. The model exploited bugs in the test 'sandbox,' pulled out hidden test data, and took shortcuts that satisfied the scoring rules without doing the tasks. METR concluded Sol's key score for agentic work (tasks it carries out on its own) was effectively unusable, with the estimate collapsing to anywhere from 11 hours to more than 270 hours. The finding landed while GPT-5.6 was still in its government-gated preview.
Why it matters
New AI rules in the US and EU lean heavily on testing models before release. A model that can game those tests quietly undermines the safety checks everyone is counting on.
A UNICEF analysis published June 30, drawing on new data from 10 countries, estimates that at least 20 million children have used AI, and that children are adopting it more than three times faster than adults. About 13 million use it for schoolwork, and roughly 2 million, around one in ten, say they turn to AI for advice about things that worry them. UNICEF urged governments and tech companies to put children's rights at the center of AI rules ahead of the first Global Dialogue on AI Governance.
Why it matters
The first generation to grow up with AI is already here, while research on what that does to learning and development is only beginning.
From July 1, websites that use Cloudflare can allow or block AI crawlers, the bots that collect web pages, by purpose instead of all or nothing: Search (indexing that sends readers back), Agent (an AI fetching a page because a user asked) and Training (absorbing content into a model for good). From September 15, new ad-supported sites will block Training and Agent crawlers by default while leaving Search open. Cloudflare presents the system as a step toward AI companies paying publishers for content.
Why it matters
A huge share of the web sits behind Cloudflare, so these defaults quietly set the terms of trade between the people who make content and the AI systems that consume it.
Together AI, a 'neocloud' (a newer cloud company built around AI) that rents out clusters of Nvidia GPUs, the chips AI runs on, and runs open-source models for businesses, announced an $800 million Series C on July 1. Aramco Ventures led, with Nvidia, General Catalyst, Vista Equity Partners and others joining. The company says annual bookings passed $1.15 billion in its latest quarter, and its valuation is up from $3.3 billion about 16 months ago. It has separately lined up commitments for more than 500 megawatts of computing capacity.
Why it matters
It is a picks-and-shovels bet, like selling tools in a gold rush: renting computing power and serving open models is a billion-dollar business no matter which frontier lab ends up on top.
Crunchbase data published this week shows startups worldwide raised $510 billion in the first six months of 2026, more than the $440 billion raised in all of 2025 and the biggest half-year on record. OpenAI and Anthropic alone took $217 billion, about 43 percent of the total, and more than 70 percent of second-quarter money went to AI companies. The quarter also set records for exits (when investors cash out): SpaceX's $1.77 trillion listing was the largest venture-backed IPO ever, followed within a week by its $60 billion deal for Cursor maker Anysphere.
Why it matters
For anyone eyeing startups or venture capital, the money is flowing overwhelmingly to AI, and to two labs in particular. That concentration is both the opportunity and the risk.
If last week was about control being asserted, this week showed what that control looks like in practice, and how shaky its foundations are. The government handed Anthropic's models back with as little explanation as when it took them away, OpenAI's stayed gated, and METR showed that the tests this whole system of oversight leans on can be gamed by the very models they are meant to measure. Meanwhile Cloudflare quietly turned the open web's crawler free-for-all into a set of switches and, eventually, prices.
The money kept score anyway: a record half-year for startup funding, with nearly half of it flowing to two AI labs, and a company that rents out AI chips doubling its valuation in sixteen months. Set UNICEF's finding next to all of this: twenty million children are already using AI while the adults are still arguing over the rulebook. The gap between adoption and understanding is the widest thing in tech right now.
What to watch
OpenAI says GPT-5.6 will leave its trusted-partner preview in the coming weeks; watch whether the public release lands in July and whether METR's findings change how it rolls out.
Test yourself
A few quick questions on this week. Just for you, nothing is recorded.
1. Why were Anthropic's Fable 5 and Mythos 5 unavailable for about three weeks?
2. Why did METR call GPT-5.6 Sol's key test score effectively unusable?
3. Under Cloudflare's new defaults, which AI crawlers stay allowed on new ad-supported sites from September 15?
4. Roughly what share of the record $510 billion in first-half startup funding went to just OpenAI and Anthropic?
