14 Jul 2026
A security firm pointed ten leading AI models at a fake company network to see how well they hack. Decoy "canary" traps flagged the intrusion before it mattered in nearly every case — and just warning the AI that traps might exist made it far more cautious. It is a striking result, from a company that sells the traps.
13 Jul 2026
Solos has announced the AirGo A6, AI smart glasses that deliberately skip the camera for a lighter, more private, everyday design. Here is what they do, how they differ from camera-equipped rivals like Meta's Ray-Bans, and the honest catch — you can't buy them yet.
13 Jul 2026
Meta has launched Muse Spark 1.1, an "agent" model built to carry out multi-step tasks across apps, and opened a paid developer API for it. Here is what it does, where it stands against Claude and GPT — and why the way Meta is shipping it is a story in itself.
12 Jul 2026
The Canadian province says it used Claude Code to review 466 million lines of government code in about 20 hours and fix the security gaps it found. Here is what actually happened, what the numbers do and don't prove, and why it matters if your own government or employer is weighing AI code review.
12 Jul 2026
Apple has sued OpenAI, its former hardware chief Tang Tan and Jony Ive's old startup, alleging a pattern of trade-secret theft aimed at OpenAI's coming AI device. The claims are unproven — but the filing shows how bitter the two companies' split has become.
12 Jul 2026
Meta launched a feature that let anyone generate AI images from a public Instagram account just by tagging it — then switched it off within days after backlash from actors' unions and agencies. The row is a lesson in "opt-out by default," and a nudge to check your own settings.
12 Jul 2026
A new preprint from the University of Maryland and Google DeepMind says the giveaway in AI-written fiction isn't its punctuation or vocabulary — it's the shape of the story. Structure alone flagged the machine about 93% of the time.
11 Jul 2026
After weeks restricted to a government-vetted few, OpenAI's GPT-5.6 — its Luna, Terra and Sol models — is now available to everyone. Here is what the three tiers are, what actually changes for you, and how to read the benchmark claims.
11 Jul 2026
Patreon has partnered with Cloudflare to block known AI training crawlers across the whole platform. Here is what that actually protects, what it doesn't, and how it fits a wider shift in who controls web content.
11 Jul 2026
OpenAI has launched ChatGPT Work, an agent that acts across your apps and files and sticks with a task for hours — and is shutting down its year-old Atlas browser to fold everything into one desktop app. Here is what it does and what to watch for.
10 Jul 2026
A shorter week than the last, but a revealing one. ChatGPT learned to talk over you, Meta's new image tool can use your face by default, and Anthropic's Cowork agent followed you onto your phone. The harder question underneath ran through half the stories: how do you tell human from machine at all? All 13 stories we covered, grouped and in plain English.
10 Jul 2026
OpenAI has replaced the model behind ChatGPT Voice with GPT-Live, which can listen and speak at the same time. It is a real step up in how natural the conversation feels — but the biggest capability claims are OpenAI's own, and the hard reasoning is still handed off in the background.