7 Jul 2026
A packed stretch: the US government asked OpenAI to gate GPT-5.6 before launch while quietly restoring a chosen few to Anthropic's Mythos, Apple raised Mac and iPad prices on an AI-driven memory shortage, and Google's own report showed its electricity use jumping 37% in a year. All 32 stories we covered, grouped and in plain English.
3 Jul 2026
Anthropic's new Claude Science app is pitched as a research workbench for scientists. But at its launch, Anthropic told reporters it plans to go further and develop medicines of its own, putting it in a three-way race with Google and OpenAI to get AI directly into the drug-discovery business.
2 Jul 2026
Google's annual electricity consumption rose 37% in 2025, its largest increase ever, driven by the AI data-centre buildout. Its own report is unusually frank: efficiency and clean-energy deals are real, but they are not keeping pace.
2 Jul 2026
OpenAI has reportedly proposed handing the US government a 5% ownership stake, and wants rival AI firms to do the same. The talks are early and unsigned, but they mark a striking shift in how Washington and the AI industry see each other.
2 Jul 2026
A UK study asked an AI to impersonate 112 real public figures from Question Time. Participants rated the AI's answers as more authentic, coherent and relevant than the politicians' own words — a finding the researchers call a "dire" warning.
1 Jul 2026
Three weeks after the US government forced Anthropic to pull Claude Fable 5 and Mythos 5, the export controls are gone and Fable 5 returns worldwide on 1 July. The more important story is what the resolution reveals: why the models were pulled, that weaker models could do the same thing, and a new, formal role for government in vetting frontier AI before it ships.
1 Jul 2026
Ask ChatGPT, Claude or Gemini for a random number between 1 and 10 and you'll almost always get 7. It's a symptom of something an award-winning study calls the "Artificial Hivemind" — the tendency of large language models to converge on a narrow band of predictable answers, and on the same answers as each other. Here's why it happens, and what to do when you actually want variety.
1 Jul 2026
Google's research assistant NotebookLM is adding Short Video Overviews — 60-second vertical clips generated from your own sources. Here's what they are, who can use them yet, and whether a bite-sized AI video of your notes is actually useful.
1 Jul 2026
A White House order wants tens of thousands of federal websites redesigned, with AI doing much of the work, by 4 July 2026. The early sites show what "vibe coding" looks like when the stakes are public — and why speed without review is the problem.
1 Jul 2026
Netflix's new Willy Wonka competition show revives Gene Wilder's voice with AI, made with ElevenLabs and cleared by his estate. Here's what's confirmed, what tool did it, and why "the family said yes" answers the legal question but not the harder one.
30 Jun 2026
The library ebook app Libby is adding a setting to filter out AI-generated content. It is more than any big rival offers — but it depends on publishers owning up to which books are AI, the very step that has failed so far.
30 Jun 2026
Anthropic has released Claude Sonnet 5, a new mid-priced model it says performs close to its flagship Opus model on many tasks — and it is now the default for free and Pro users. The capability-for-price step is real, but the headline numbers are Anthropic's own, and a quiet tokenizer change softens the "cheaper" story.