Claude Fable 5 is back online after a government-forced shutdown, AI is now the top reason for layoffs three months running, and the EU's AI enforcement clock is down to seven weeks.
Claude Fable 5 restored, after the first government-forced AI shutdown
Anthropic restored access to its most powerful model, Fable 5, after a six-day blackout. The US Department of Commerce had issued an export control directive on June 12 ordering Anthropic to suspend access for all foreign nationals (including its own employees) over a reported jailbreak vulnerability. Anthropic disputed the severity of the finding, and senior staff negotiated directly with White House officials to restore access. Mythos 5, a sister model, was also taken down and is still under review.
Why it matters: This is the first documented case of a government forcing a major AI company to pull a commercial model from hundreds of millions of users. Whatever you think of the reasoning, the precedent is now set: states can switch off AI. That changes the risk calculus for every business built on top of these models.
AI is now the top reason for layoffs, three months in a row
New data shows 363 rounds of layoffs in 2026 have eliminated nearly 150,000 jobs, about 974 people a day, running 44% faster than last year at the same point. For the third straight month, companies cited AI automation as the primary driver. Meta alone cut 8,000 roles in May. The job cuts span tech, media, legal, and support functions where AI agents are now completing tasks that previously required full-time staff.
Why it matters: This is no longer a warning about the future, it is the present. If your job involves repetitive information work, the question of what to do next is worth taking seriously now, not in two years.
Microsoft's MAI family: seven new models, including one that codes better than most engineers
Microsoft launched seven new AI models at Build 2026 under the MAI brand. Highlights include MAI-Image-2.5 (now ranked second globally on image quality leaderboards), MAI-Thinking-1 (a reasoning model scoring 97% on AIME math and 53% on software engineering benchmarks), and MAI-Code-1-Flash, a five-billion-parameter coding model that achieves 51% on SWE Bench Pro while running efficiently inside VS Code and GitHub Copilot. The models are rolling out on Azure Foundry, OpenRouter, and Fireworks.
Why it matters: Microsoft is no longer just reselling OpenAI. It now has its own competitive model stack, which gives developers options and pushes prices down. MAI-Code-1-Flash in particular is a strong tool for students learning to code with AI assistance.
Half of all businesses have now been hit by deepfake fraud
A new study finds that 50% of companies have encountered AI-driven deepfake fraud in some form, most commonly fake video job interviews where candidates use real-time face-swapping to impersonate someone else. Hiring managers and security teams report the problem jumped sharply in Q1 2026. AI-generated voice calls impersonating executives to authorise payments are also rising. The tools used are consumer-grade and freely available.
Why it matters: If you are hiring, you need a verification step that a video call alone cannot provide. If you are job-seeking, you are entering a market where your identity needs to be demonstrably yours. Both sides of the table are affected.
EU AI Act: full enforcement is 42 days away
The EU AI Act becomes fully applicable on August 2, 2026, seven weeks from today. An "AI omnibus" simplification package was agreed in May to reduce compliance burden for smaller companies, but high-risk AI systems in healthcare, education, and employment still face strict rules. The EU also reached a political agreement to streamline parts of the Act after significant industry pushback. Non-compliance fines can reach 3% of global annual turnover for most violations, rising to 6% for the most serious breaches.
Why it matters: If your company uses or builds AI tools and operates in Europe (or serves European users) you have a hard countdown. The August 2 date is not moving.
India defines AI-generated content in law, and trains 15,000 media creators
India's 2026 IT Rules amendment now formally defines "Synthetically Generated Information" (SGI), audio, visual, or audio-visual content created or altered by AI to appear real. Platforms must label and in some cases remove such content. Separately, the Indian government announced a programme to train 15,000 news creators and media professionals on responsible AI use, in partnership with Google and YouTube. The Rajasthan AI-ML Policy 2026 was also unveiled at a regional conference, with a focus on AI in agriculture and public services.
Why it matters: India is building its AI legal framework piece by piece, state by state. The SGI definition is significant, it gives regulators a legal handle on deepfakes for the first time. The media training programme is practical and immediate.
Who controls access to the most powerful AI tools?
The Fable 5 shutdown showed that governments can now flip a switch on commercial AI at scale. Anthropic got its model back after six days, but Mythos 5 is still offline. The questions this raises are not hypothetical: what happens to businesses built on API access? What happens to users in countries that get caught in a geopolitical directive? And if one government can do this, what stops others? The next 90 days of AI governance will matter more than the previous 90 launches.
Suno raises $400M for AI music, Supabase hits $10.5B, and Canada bets on AI for 90,000 jobs
AI music platform Suno closed a $400M Series D at a $5.4B valuation, making it one of the best-funded creative AI companies. Developer infrastructure firm Supabase raised $500M at a $10.5B valuation. And Canada launched AI for All, a national strategy promising 90,000 AI-related jobs and work placements for young Canadians, one of the first national programs to frame AI primarily as a jobs opportunity rather than a threat.