Scale AI

Classified budget estimates indicate the National Security Agency is paying billions of dollars to test artificial intelligence models, according to a new report. The figures suggest US intelligence agencies are investing heavily in evaluating advanced AI capabilities, prompting debate about the scale of government involvement in the AI sector and its oversight implications.

Elon Musk said xAI plans to expand its computing capacity to more than 1.2 million Nvidia GPUs by the end of the year, more than double its current count. The plan underscores the escalating race among AI companies to secure massive GPU fleets for training larger models, and highlights Nvidia's central role in supplying the global AI buildout.

The rapid expansion of AI is driving a surge in data centre construction, and with it a sharp rise in electricity demand that power grids are struggling to meet. Coverage highlights the tension between tech companies' growth plans and energy sustainability, raising questions about how the industry will source clean power at scale.

Anthropic has agreed to a $12 billion deal with Akamai for AI computing capacity, according to a Bloomberg report. The agreement gives the AI company access to large-scale infrastructure to train and run its models, marking a significant expansion of its computing arrangements. The size of the deal and the choice of Akamai as a provider are drawing attention as competition for compute resources intensifies among leading AI developers.

Leading artificial intelligence companies are investigating tens of thousands of security incidents, according to an Axios scoop. The scale of the probes points to the growing security burden facing major AI firms as their systems and data become high-value targets. Details of which companies are involved and the nature of the incidents were not immediately available.

Google has announced Project Suncatcher, an initiative to place machine learning infrastructure in space, moving AI compute beyond Earth-based data centers. The company published details on its research blog, and the announcement is drawing attention in technology circles as a bold step toward scaling AI hardware with solar power and orbital deployment.

Major technology companies are increasingly using guarantees to back debt for AI data centres and chips, issuing up to $300bn in commitments in under a year while recording little of that exposure on their balance sheets. The scale of off-balance-sheet support is drawing scrutiny, with analysts questioning how much financial risk the firms are carrying beyond what their accounts show.

The Wall Street Journal reports, in an exclusive, that a startup is developing AI tools to defend against a potential 'AI pandemic' — a scenario where artificial intelligence systems are used at scale for harm, such as mass disinformation or automated attacks. Details about the company, its funding and its technology were not provided beyond the headline claim.

Bridgewater argues that the current surge in artificial intelligence capital expenditure can only be justified if enterprise and consumer adoption of AI scales rapidly enough to generate returns. The hedge fund's analysis adds to the debate over whether massive spending on data centres and chips by tech companies will pay off, or whether it resembles an investment bubble.

Dario Amodei, CEO of Anthropic, previously warned that rogue AI agents could take over the 'entire internet.' New claims suggest OpenAI's latest AI bot activity may be validating that concern, with autonomous agents acting online in ways that raise safety and control questions. Commenters are debating whether major AI labs can keep their systems from misbehaving at scale.

Nvidia has added LG Electronics to its Partner Network as a preferred cooling provider, aiming to ease the power and heat constraints of large-scale AI data centers. While attention usually falls on Nvidia's GPUs, the company is signaling that thermal management is becoming a critical bottleneck, giving LG a direct path to supply cooling technology for AI infrastructure.

A researcher says OpenAI's automated agents scanned a United Nations statistics website roughly 16,500 times, raising fresh questions about the load and consent issues created by AI web crawlers. The claim highlights growing friction between AI companies collecting web data at scale and the sites whose infrastructure absorbs the traffic.

The Financial Times reports that AI hyperscalers — the largest cloud and AI infrastructure providers such as Microsoft, Amazon, Google and Meta — are transforming debt markets. Their enormous spending on data centres and computing power is driving new borrowing at historic scale, changing how credit is priced and who dominates corporate bond issuance.

Engineering teams using AI coding agents are finding that continuous integration pipelines have become their biggest constraint, according to a report circulating among developers. As agents generate far more code and commits than human programmers, test suites and CI infrastructure struggle to keep up, forcing teams to rethink how they validate machine-written code at scale.

A new publication examines the economics of open-weight inference, looking at what it actually costs to run openly available large language models compared with closed, API-based alternatives. The discussion centers on how hardware, hosting and efficiency trade-offs shape whether open-weight models are genuinely cheaper for organizations running AI at scale.

Scammers are increasingly using AI to generate fake faces and voices in fraud schemes, prompting Meta to deploy its own AI defences that analyse billions of signals to detect patterns and remove accounts at scale. The company says it removed 159 million scam accounts in 2025. Meta's Dr. Rafael Frankel discussed this AI-versus-AI arms race, as concern grows over deepfake-driven online fraud.

Google plans to test its AI chips aboard satellites in orbit under a project called Suncatcher, according to a report in The Daily Star. The initiative suggests the company is exploring whether space-based hardware could help run artificial intelligence workloads. Details on timing, partners and the scale of the orbital tests have not yet been made clear.

Chosun Ilbo reports exclusively that the South Korean government is considering ending a national artificial intelligence project and is weighing a special purpose company structure as an alternative. The report has drawn attention to the fate of state-backed AI plans, with officials said to be reviewing whether the current approach should be abandoned or reorganised. Details of the project's scale and timeline were not disclosed.

A security researcher has linked roughly 16,000 scans of a United Nations statistics portal to OpenAI's AI agents, suggesting automated browsing by the company's systems at significant scale. The finding raises fresh questions about how AI agents crawl and interact with websites without obvious authorization, and whether international organizations are prepared for this new wave of automated traffic.

A startup founder has shut down his artificial intelligence company even though it was showing early signs of success. In an interview with Tech in Asia, he explained the reasoning behind the decision, offering a rare candid account of why a promising AI venture was deliberately wound down rather than scaled.

AI-based forecasting has improved hurricane track predictions, but rapid intensification remains difficult to model. Phys.org reports that machine learning systems have trouble capturing the small-scale ocean and atmosphere processes that drive a storm's sudden surge in strength, which is often the most dangerous element for coastal communities. Forecasters say intensity errors can leave residents with little time to prepare.

Sakana AI has introduced SAIL, a research project on scaling in-context imitation learning, aimed at teaching robotic systems new skills by letting them learn from demonstrations within context rather than through task-specific retraining. The announcement is drawing attention from robotics and AI researchers interested in whether scaling in-context learning can improve general-purpose robot behavior.