Fable 5 returns under government control, Altman offers 5% of OpenAI, and AI starts to look conscious

Peter Diamandis, Dave Blundin, and Alex frame this week’s AI news as a shift from product competition to strategic infrastructure. Fable 5’s return after government-linked restrictions becomes a case study in how frontier models may now carry national-security obligations, while the panel debates whether labs can truly identify users, inspect prompts, and block sophisticated misuse.

Main signals

• Frontier AI labs are beginning to look like semi-public institutions with security duties.
• User identity, prompt monitoring, and model-internal inspection are becoming separate but connected governance problems.
• Anthropic’s work on a Claude « global workspace » suggests interpretability may expose structures that resemble cognitive machinery.
• Sam Altman’s call for a US-led international governance forum raises the question of whether China participates or the world splits into AI blocs.
• Ramp/Ravelio data points to a more optimistic employment story: deep AI adopters are hiring more, not less.

Why it matters

The practical theme is sovereignty. Companies and governments want control over models, compute, data, prompts, and the value created by AI systems. Palantir and Nvidia’s sovereign AI architecture is presented as a response to customers who do not want their critical workflows dependent on external token vendors or export-control risk.

What to watch

• A potential East-West split in superintelligence access and model exports.
• Mechanistic interpretability becoming a trust and alignment layer.
• AI-designed chips and RF circuits accelerating the innermost loop of recursive improvement.
• Patent law pressure as AI systems generate increasingly valuable inventions.

Source

• Chaîne: Peter H. Diamandis
• Vidéo source: https://www.youtube.com/watch?v=XCunMF6frio