Hello AI Enthusiast,
This week showed how much control governments now have over AI releases. Washington lifted its export ban on Anthropic's Fable 5 and Mythos 5, only to put its own limits on OpenAI's brand new GPT-5.6 family. Meanwhile, Anthropic opened a dedicated workspace for scientists, and OpenAI published a report claiming Codex has quietly become the default way people work. Let's get into it.
The Big Picture 🔊
US lifts Fable 5 ban as GPT-5.6 launches to a limited group
Anthropic's Fable 5 and Mythos 5 returned to customers after a two-week export-control suspension, once US regulators signed off on new safeguards and got commitments on ongoing monitoring. Meanwhile, OpenAI launched its GPT-5.6 family, Sol, Terra, and Luna, with Sol showing stronger coding and cybersecurity capabilities. Only a small group of trusted partners can access it so far, at the government's own request, ahead of a wider release in the coming weeks.
|
||
|
Companies hype up how dangerous their models are, but that backfires once a government believes them and shuts it down. That risk is now permanent for IPO filings. It also exposes a deeper problem for Europe, which built its whole strategy around using US models while regulating them. If access can be cut off, Europe ends up with nothing to regulate and nothing of its own. |
Anthropic launches Claude Science for researchers
Anthropic launched Claude Science, a dedicated workspace built on existing models like Opus 4.8, connected to over 60 research tools including PubMed and protein viewers. A coordinating assistant delegates to sub-agents, while a separate reviewer checks citations and calculations. Data can stay on the researcher's own infrastructure. Anthropic is also starting its own pre-clinical drug programs. Now in beta for Pro, Max, Team, and Enterprise users.
|
||
|
Anthropic keeps finding narrow, specific gaps, and science looks like a market nobody has served well yet. The provenance feature is what makes it useful beyond convenience, since a wrong source in a published paper can get it retracted and set back a career. Whether a generalist company can keep doing this for every industry is a separate question. |
OpenAI report claims Codex is now the company's default too
OpenAI published research claiming its coding tool Codex has become the default AI tool across every department, not just engineering, with employees generating over 85% of their output tokens through Codex rather than ChatGPT. Non-developer users grew more than a hundredfold since August 2025, and by May 2026 a quarter of Codex requests involved tasks estimated to take a person more than eight hours. The numbers come entirely from OpenAI's own data.
|
||
|
A company grading its own homework and calling it a landmark study is still an ad, whatever the numbers say. What's worth paying attention to is the direction: Codex is being pitched as the default tool for non-technical staff, not just developers. Even we don't always know whether a task calls for chat or an agent tool like Codex. |
OpenAI's report shows employees leaning harder on tools like Codex, but having access to a tool isn't the same as knowing when to use it. Our Corporate AI Training helps your team build that judgment, so people pick the right tool instead of guessing.
Bits and Bobs 🗞️
Anthropic launched Claude Sonnet 5, a cheaper, more agentic model that matches Opus 4.8 on some tasks while improving safety.
Google released Nano Banana 2 Lite and Gemini Omni Flash, two lightweight models for fast AI image generation and video animation.
Google introduced computer use in Gemini 3.5 Flash, letting the model click and type its way through interfaces on its own, backed by new safety training for autonomous action.
OpenAI and Broadcom unveiled Jalapeño, a custom AI chip built to make running large language models faster and cheaper at scale.
Notion launched a Developer Platform called Workers, letting AI agents connect to external tools like CRMs and support systems and run custom logic directly inside Notion.
From Our Founder’s Channels 🤳
Gianluca and his AI predictions:
@gianluca.mauro I predicted this 2 weeks ago 🔮 #Ai #learnontiktok #business
LOLgorithms 😂
When you're afraid of the US government's second thoughts.
That's a wrap on our newsletter! Before you go, here’s a quick recap of our offerings:
AI Academy Membership: Get 12 months of access to all our cohort-based programs and on-demand courses.
AI Agent Bootcamp: Accelerate processes and solve business problems by building AI Agents, without coding.
Practical Introduction to ChatGPT: A free course on using ChatGPT confidently, understanding its workings, and exploring its potential.
Customized Corporate Training: Equip your team with the skills they need to unlock the potential of AI in your business.
Catch you next week! 👋
