🛡️ Bessent, Powell warned bank CEOs about Anthropic model risks, sources say ↗
Anthropic’s new Mythos model sparked enough concern that U.S. Treasury Secretary Scott Bessent and Fed Chair Jerome Powell reportedly called bank CEOs into an urgent briefing. That is not a routine AI launch - it feels closer to a fire drill in a server room.
The issue is cyber risk. Anthropic said Mythos can identify and exploit weaknesses across major operating systems and web browsers, which is why the company held back a broad release and kept consulting government officials instead. Sobering and significant, both can be true at once. (Reuters)
☁️ CoreWeave strikes AI cloud deal with Anthropic, shares surge ↗
CoreWeave signed a multi-year deal to supply Anthropic with more cloud capacity for Claude workloads, and investors welcomed it - the stock climbed more than 13%. The core story is straightforward: everyone wants compute, supply remains tight, and the pipes keep getting pricier.
The capacity is expected to come online later this year, with room to expand. It also helps CoreWeave diversify beyond heavy dependence on Microsoft, while Anthropic keeps stacking supply deals after its Google-Broadcom arrangement and its reported interest in designing chips of its own. Bit by bit, the AI arms race is starting to resemble an infrastructure auction. (Reuters)
🌍 South Africa unveils draft AI policy, proposes new institutions and incentives ↗
South Africa published a draft national AI policy that is trying to do two jobs at once - accelerate adoption and place genuine governance around it. The plan includes a National AI Commission, an AI Ethics Board and an AI Regulatory Authority, which sounds bureaucratic, certainly, but also increasingly inevitable.
There is an economic case behind it as well. The draft raises tax breaks, grants and subsidies for startups and smaller firms, while pushing investment into supercomputing and network infrastructure. At the same time, it flags a concern many countries quietly share - leaning too heavily on foreign cloud and hardware could leave sensitive data exposed and make national AI plans feel somewhat hollow. (Reuters)
🌫️ How the AI boom derailed clean-air efforts in one of America's most polluted cities ↗
One of the darker AI stories yesterday was not about models at all - it was about electricity. Reuters traced how U.S. policy rollbacks meant to support AI-driven power demand are keeping dirtier coal generation alive around St. Louis, where activists had expected tighter soot rules to force cleaner outcomes at last. Bleak, by any measure.
The piece says local groups see the AI buildout as a direct threat to air quality, especially in predominantly Black communities already carrying the heaviest burden. So the gleaming AI future keeps asking for “just one more data center,” and somewhere else the cost arrives as soot, asthma and higher health bills. A very old problem for such a futuristic industry. (Reuters)
⚛️ Big Tech puts financial heft behind next-gen nuclear power as AI demand surges ↗
Meta, Amazon and Google are putting serious weight behind advanced nuclear projects as they chase power for AI data centers. The appeal is fairly obvious - small modular reactor developers get funding and, crucially, long-term buyers, which makes the whole pitch look less like science-fair optimism and more like bankable infrastructure. At least for now.
Reuters notes Meta is backing TerraPower and Oklo projects, Amazon is working with X-energy, and Google has a deal with Kairos Power. Demand from AI is making these reactor bets feel less optional and more like the industry’s backup generator for the next decade - except the backup is starting to look like the main plan. (Reuters)
FAQ
Why did Anthropic’s Mythos model trigger concern from bank regulators?
Anthropic said Mythos could identify and exploit weaknesses across major operating systems and web browsers, which immediately raised cyber risk concerns. That appears to explain why U.S. officials reportedly briefed bank CEOs instead of treating it like a routine model launch. In practice, the concern is less about publicity and more about whether powerful AI systems could speed up offensive security capabilities.
What does the CoreWeave and Anthropic deal mean for the AI cloud market?
The deal underscores how valuable AI cloud capacity has become as demand for model training and inference continues to rise. CoreWeave gains a major customer relationship beyond Microsoft, while Anthropic secures additional infrastructure for Claude workloads. More broadly, it suggests the AI cloud market is turning into a contest over scarce compute, long-term supply, and who can scale the fastest.
What is South Africa trying to achieve with its draft AI policy?
The draft policy seeks to balance faster AI adoption with stronger governance. It proposes new bodies such as a National AI Commission, an AI Ethics Board, and an AI Regulatory Authority, while also supporting startups through incentives like grants, subsidies, and tax breaks. The policy also reflects a strategic concern that overreliance on foreign cloud and hardware could weaken national control.
How is the AI boom affecting air quality and local communities?
The article suggests that rising electricity demand from AI infrastructure can keep older, dirtier power sources operating for longer. In the St. Louis area, that has raised fears that local communities, especially predominantly Black neighborhoods already carrying heavier pollution burdens, could face worsening air quality. The broader takeaway is that AI growth can shift environmental costs onto places far from the data centers themselves.
Why are Meta, Amazon, and Google investing in advanced nuclear power for AI?
AI data centers need large, reliable electricity supplies, and advanced nuclear projects offer one potential long-term answer. By backing reactor developers and signing power agreements, big tech companies can help make those projects more financially viable. The move also shows that for many firms, energy planning is no longer separate from AI strategy; it is becoming part of the core infrastructure roadmap.
What ties these AI stories together into one bigger trend?
They all point to AI becoming an infrastructure and governance issue, not just a software story. Cybersecurity risks, cloud supply constraints, national policy, pollution tradeoffs, and nuclear financing are now part of the same conversation. That matters because the next phase of AI competition appears to depend as much on power, regulation, and resilience as on model performance.