David Sacks says Anthropic declined to patch Fable 5 before the export control directive, while Microsoft walked from a $3 billion Oracle commitment as compute scarcity bites.
Sacks's claim recasts last week's jailbreak from a technical lapse into a governance dispute, with Anthropic offered a choice before the export control directive.
The agreement extends the ceasefire by 60 days and reopens the Strait of Hormuz, pulling oil prices lower and lifting Indian equities and the rupee.
Microsoft is adding AWS capacity for GitHub while abandoning a $3 billion Oracle commitment, signaling compute scarcity forces pragmatic, even adversarial moves.
David Sacks, Co-Chair of the President's Council of Advisers on Science and Technology, revealed on June 13 that the Trump administration offered Anthropic a choice before the export control directive: patch the jailbreak vulnerability or voluntarily de-deploy Fable 5. According to Sacks, CEO Dario Amodei refused...
Continue reading arrow_outwardDavid Sacks, co-chair of the President's science and technology council, revealed that Anthropic declined to deploy a fix for its Fable 5 model, escalating tensions over last week's jailbreak fallout.
The US and Iran signed an agreement to reopen the Strait of Hormuz, easing a chokepoint that had rattled global energy markets.
Sentiment remains pinned at Extreme Fear despite the Hormuz breakthrough, with falling oil only briefly lifting Indian equities and the rupee.
Microsoft is leaning on rival AWS for GitHub AI capacity while walking away from a $3 billion Oracle cloud deal, a sign that even hyperscalers cannot self-supply compute.
Sacks's claim that Anthropic refused a fix recasts last week's jailbreak from a technical lapse into a governance dispute. With a separate rumor of Claude Sonnet 4.8 now confirmed false, Anthropic faces scrutiny over both its safety posture and its product roadmap heading into a planned IPO.
Microsoft tapping AWS for GitHub workloads while abandoning a $3 billion Oracle commitment signals that compute scarcity is forcing pragmatic, even adversarial, partnerships. The moves suggest internal capacity for AI coding tools is tighter than the company's public messaging implies.
The US-Iran deal to reopen the Strait of Hormuz removes a major geopolitical risk and pulled oil prices lower, boosting Indian markets. Yet the broader mood stays at Extreme Fear, indicating investors see structural risks beyond the Middle East, including AI sector volatility.
Microsoft Pays Its Biggest Rival
Read full analysis arrow_outwardDavid Sacks, Co-Chair of the President's Council of Advisers on Science and Technology, revealed on June 13 that the Trump administration offered Anthropic a choice before the export control directive: patch the jailbreak vulnerability...
The weeks-long rumor of a Claude Sonnet 4.8 release appears to be false, traced to a misread leak. Anthropic's current Sonnet model remains Claude Sonnet 4.6. The clarification ends speculation that opened a release window earlier this week.
Microsoft is adding Amazon Web Services capacity to keep GitHub running after an AI-driven surge in coding activity strained the platform, per a Business Insider report citing two people familiar with the plans. The move cuts against...
Microsoft was recently in talks to lease Oracle cloud infrastructure but walked away over security and compliance concerns, according to people familiar with the matter. One source said the deal could have been worth more than $3...
OpenAI published research describing a near-autonomous AI chemist that improved a challenging medicinal chemistry reaction, showcasing AI's potential to accelerate chemical synthesis. The company separately introduced LifeSciBench, a...
A June 4 congressional discussion draft, the Great American AI Act, would create federal transparency and reporting duties for large frontier AI developers and establish CAISI as a technical hub for AI testing, standards, and...
Databricks announced Genie One, a governed AI coworker that turns business context into actions through agentic automation and a data context ontology. The product targets enterprise AI governance concerns, aiming to give organizations...
The US reportedly reached an agreement with Iran that extends the ceasefire by 60 days and reopens the Strait of Hormuz, ending the US blockade of Iranian ports. A formal signing of the memorandum of understanding is expected in...
Indian equities and the rupee rose after crude prices fell on reports of the US-Iran deal and the prospect of easier oil flows through the Strait of Hormuz. Improved investor sentiment reflected expectations of a lower Middle East risk premium.
OpenAI is strengthening its Codex coding agent through the Ona acquisition, adding secure, persistent cloud workspaces so agents can run multi-hour or multi-day tasks without losing context. Use it for long-running refactors, background task execution, and stateful agentic coding in isolated environments.
Try it arrow_outwardLemonade is a local AI runtime that began on AMD hardware and now also runs on NVIDIA GPUs via CUDA, giving teams cross-vendor flexibility for on-device inference. Deploy it when you want to run local models without locking your stack to a single GPU vendor.
Try it arrow_outwardNVIDIA published a quantized DiffusionGemma 26B model (diffusiongemma-26B-A4B-it-NVFP4) on Hugging Face, packaged for fast local inference on NVIDIA hardware. Builders can experiment with diffusion-based text generation locally without paying per-token API fees.
Try it arrow_outwardTAAFT is a large AI tool directory with personalized search, real-time trends, ratings, and reviews to help you find the right tool for a specific task instead of guessing. It is free with paid options available, useful when scoping automation or vetting new tools quickly.
Try it arrow_outwardSpin up a persistent workspace. Try OpenAI Codex with persistent cloud workspaces to keep long-running coding agents stateful across sessions instead of re-priming context each run.
Run local inference cheaply. Test the AMD Lemonade SDK to offload coding and chat workloads to local hardware, hedging against the compute scarcity now forcing Microsoft to shop AWS for GitHub.
Benchmark a smaller diffusion model. Evaluate NVIDIA DiffusionGemma 26B NVFP4 for quantized generation, and confirm you are building against Claude Sonnet 4.6 since the 4.8 rumor is false.
Sacks's Fable 5 disclosure turns a model exploit into a policy fight, and Microsoft ditching $3 billion of Oracle commitment shows even giants are scrambling for capacity. Builders should treat single-vendor compute and single-vendor governance as real risks. Diversify your inference across clouds and local SDKs, and verify model versions yourself since the false Sonnet 4.8 rumor proves leaks lie. Markets are jittery and crypto sits in extreme fear, so ship lean and stay portable.
AI analysis, world news, markets, and tools. One briefing, delivered free.