Fable 5 is back, Sonnet 5 is now everyone's default, and the UN warned the window to fairly govern AI is closing faster than governments are moving.
Anthropic's biggest week: Fable 5 returns after a 19-day ban, Sonnet 5 goes live for everyone
Fable 5 (Anthropic's most powerful model) was restored on July 1 after US export controls that pulled it offline on June 12 were lifted on June 30. The ban was triggered by a jailbreak prompt discovered by Amazon researchers; Anthropic argued the controls were unjustified because every other major model tested was vulnerable to the same exploit. The same week, Anthropic launched Claude Sonnet 5 as the new default for Free and Pro users, with a 1-million-token context window, stronger agentic coding performance (63.2% on benchmarks, up from 58.1%), and introductory pricing of $2 per million input tokens through August 31.
Why it matters: If you use Claude, you've quietly been upgraded this week. But the bigger story is what Fable 5's 19-day shutdown reveals: the most powerful AI tools can be switched off overnight by government order. That's not a warning about the future, it happened last month.
UN: "The window to govern AI fairly is closing, and inequality will fill the gap"
UN experts published a stark assessment on July 2: governments are falling behind the pace of AI deployment, and without urgent coordinated global action, AI will entrench and deepen existing inequality. The report flagged that AI development is overwhelmingly concentrated in a small number of wealthy countries and a handful of companies, with OpenAI and Anthropic alone accounting for 43% of all global startup funding in the first half of 2026. Most of the world has little input over systems that increasingly affect them.
Why it matters: This is the kind of warning that tends to arrive just before the window actually closes. For students and professionals in the Global South (including India) it's a signal that participation in shaping this technology, not just adopting it, matters now more than ever.
AI regulation is no longer coming, it's here. Colorado's AI Act took effect June 30.
Colorado's AI Act became enforceable on June 30, making it the first US state law to place binding legal obligations on companies that develop or deploy AI in high-stakes decisions, including hiring, lending, and healthcare. Companies must now assess and actively work to prevent algorithmic discrimination, or face real legal consequences. Separately, the EU's AI Act compliance deadline for high-risk AI systems falls on August 2, 2026 (five weeks away) covering data quality, transparency requirements, and human oversight mandates.
Why it matters: Regulation is no longer a distant scenario to plan for. If you're building an AI product, using AI to screen job applicants, or working in a company that does either of those things, the legal environment changed this week. The EU deadline in August is the bigger global one to watch.
AI-powered deepfake scams have become a $68 billion crisis, and anyone can be targeted
A new survey published July 1 found that AI deepfakes are driving a $68 billion scam epidemic in the United States alone. The technology has crossed a threshold: deepfakes no longer have the visible glitches that once gave them away, and the tools to create them are now accessible via a smartphone. Documented scam types include fake video calls from "colleagues" and "bosses" (one UK engineering firm lost $25 million to a single fake video call), AI-generated job candidates getting through interviews, and fake investment opportunities featuring real people's faces and voices.
Why it matters: "Seeing is believing" no longer applies. If you receive a video or voice message asking for urgent action (money, access, sensitive information) verify through a completely separate channel before acting. This applies at work and personally.
India puts AI in schools from Grade 3, while IT job pressure grows at the other end
India announced that an AI curriculum will roll out across all schools starting Grade 3 from the 2026-27 academic year, aligned with the National Education Policy 2020. The Ministry of Education is establishing a Centre of Excellence in AI for Education with Rs.500 crore in dedicated funding. At the same time, analysts at Bernstein have formally warned the Prime Minister of a deepening employment crisis in India's IT sector as AI automates high-skill work, with 9.2 lakh AI workers in the country, but 84% reporting they feel unprepared. India's proposed AI Economic Council will aim to coordinate between infrastructure investment and workforce readiness.
Why it matters: India is investing in the right generation, the one starting school now. The harder question is what happens to the current generation of IT workers caught mid-career. The gap between policy announcements and actual reskilling support is where the real risk sits.
Global AI startup funding hit $510 billion in the first half of 2026, a record with no precedent
Crunchbase data shows global startup investment reached $510 billion in H1 2026, surpassing the entire $440 billion invested across all of 2025. AI dominated: nearly 88% of AI-related funding went to US companies, and OpenAI and Anthropic alone took $217 billion: 43% of all startup investment globally. Outside the frontier model race, significant capital is now flowing into AI infrastructure, healthcare AI, defense-tech, and robotics. Anthropic is now valued at $965 billion following its $65 billion Series H close in May.
Why it matters: The financial concentration in AI is without historical precedent. For students and early-career professionals, this is a signal about where jobs, opportunities, and economic activity are being built for the next decade, and also a reminder of the inequality the UN flagged: most of that money is flowing to one country.
The race between AI deployment speed and the ability to govern it
Colorado's AI law is now in force. The EU's high-risk AI deadline is August 2. The UN says the governance window is closing. And in the same week, Anthropic launched two models, global startup funding hit half a trillion dollars in six months, and the most powerful AI available was pulled offline and reinstated by a single government decision. This tension (deploy faster versus govern this properly) is the defining contest of 2026. What gets decided in the next 12 months will shape the rules of AI for a decade. It's worth paying attention even if you're not a policymaker.
OECD releases global AI literacy framework + OpenAI launches Flex Processing + Willow.ai brings AI to business calls
The OECD published an AI literacy framework titled "Empowering Learners for the Age of AI" for primary and secondary schools worldwide, a useful reference for educators and anyone thinking about curriculum. OpenAI launched Flex Processing, letting developers run less time-sensitive API tasks at a lower cost by letting OpenAI schedule them in off-peak windows, useful for budget-conscious builders. And Willow.ai released Atlas-1, a speech-to-text platform designed specifically for business voice calls (sales, support, and compliance) a signal that voice AI is maturing from transcription into a structured business tool.