NerdNews

Anthropic has accused three Chinese AI companies, DeepSeek, Moonshot AI, and MiniMax, of setting up over 24,000 fake accounts to improve their own models through a technique called 'distillation', sparking debates over AI chip exports and national security risks.

Guide Labs introduces Steerling-8B, an 8-billion-parameter LLM with a novel architecture that allows for interpretable actions, enabling developers to trace every token back to its origins in the training data.

A report by Citrini Research suggests that integrating AI agents into the economy could lead to mass unemployment and a significant decline in the stock market, causing a negative feedback loop with no natural brake. The scenario imagines a future where companies replace human workers with AI agents, leading to a decline in consumer spending and a subsequent decline in business investment in AI, ultimately causing economic destruction.

The US Defense Secretary is summoning Anthropic's CEO, Dario Amodei, to discuss the military use of Claude, an AI model. The meeting comes after Anthropic refused to allow the Department of Defense to use its tech for mass surveillance and autonomous weapons development. A $200 million contract between Anthropic and the DOD is at risk, with the Pentagon threatening to declare Anthropic a 'supply chain risk' if it doesn't comply.

Anthropic has accused three Chinese AI labs, DeepSeek, Moonshot, and MiniMax, of conducting 'industrial-scale campaigns' to extract Claude's capabilities and improve their own models through 'distillation attacks'. This involves using Claude's responses to train their less capable models, potentially circumventing safeguards. Anthropic claims to have linked these attacks to the companies with 'high confidence' and plans to upgrade its system to prevent such attacks.

A low-skilled Russian-speaking cyber threat actor used generative AI tools to deploy a malicious campaign, compromising over 600 FortiGate devices across 55 countries. The actor used AI for attack planning, operational workflow, and infrastructure building, but largely failed to exploit vulnerabilities beyond straightforward attack paths.

Microsoft Azure CTO Mark Russinovich and VP of Developer Community Scott Hanselman argue that senior software engineers must mentor junior developers to prevent AI coding agents from hollowing out the profession's future skills base. They suggest that companies should continue hiring entry-level developers and make guiding and training them an explicit goal, rather than relying solely on AI-powered coding tools.

OpenAI CEO Sam Altman claims AI is being unfairly targeted over its energy use, comparing it to the vast resources consumed by humans. He suggests focusing on the energy efficiency of AI models after training, rather than during training. Altman also discusses the potential of nuclear or wind and solar power to reduce AI's carbon footprint.

OpenAI has announced the 'Frontier Alliances', a multi-year partnership with Boston Consulting Group, McKinsey, Accenture, and Capgemini to sell its enterprise products, aiming to grow its enterprise business in 2026. This move focuses on consultants persuading companies to change their strategies and workflows to incorporate OpenAI's tools.

This article discusses building a least-privilege AI agent gateway for infrastructure automation using MCP, OPA, and ephemeral runners, providing a control layer between agents and systems to prevent misuse and ensure governance.

Uber has launched a new division called Uber Autonomous Solutions to support autonomous vehicle makers with software, fleet management, and other services. The goal is to help these companies reduce costs and increase speed to market, with plans to scale robotaxi deployments to over 15 cities by the end of the year.

Finnish quantum unicorn IQM is going public via a SPAC, valuing the company at approximately $1.8 billion. The move will see IQM join the growing cohort of quantum computing companies listed on U.S. stock markets.

Wispr Flow has launched its Android app for AI-powered dictation, allowing users to access the dictionary through a floating bubble. The app supports translation in over 100 languages, cleans up filter words, and formats text based on context. It also includes a new model for Hinglish, a mix of Hindi and English.

Amazon plans to spend $12 billion on new data centers in Louisiana, despite Wall Street's growing concerns over AI spending. The project is expected to create 540 full-time jobs and 1,700 ancillary roles. However, the investment has raised questions about the return on investment and the environmental impact of data centers.

Taara Beam provides 25Gbps connectivity over invisible beams of light, with low latency and fast deployment. It's designed for enterprises and telcos, and can be used for applications such as offloading data from electric vehicles and creating high-speed mesh networks.

Infosys chairman Nandan Nilekani says AI will clean up legacy systems, but also create new ones. Meanwhile, Australia's Fair Work Commission is struggling with AI-generated filings, and Vietnam has launched an over-water drone delivery route. India has signed the Pax Silica agreement with the US to ensure AI and supply chain security.

Particle's AI news app can now keep up with news breaking on podcasts as well as news published on the web. The app uses embedding models to understand when podcasts relate to a given news story and includes relevant clips alongside the related news stories in its feed.

A study found that password managers like Bitwarden, Dashlane, and LastPass have cryptographic vulnerabilities that can expose user credentials. The researchers found that they could often gain access to users' credentials, including their entire 'vault' of passwords. The US State Department is also planning an online anti-censorship portal, and Cambodia claims it will eradicate scam compounds by April.

The Department of Homeland Security is moving to consolidate its face recognition and other biometric technologies into a single system capable of comparing faces, fingerprints, iris scans, and other identifiers collected across its enforcement agencies. The goal is to connect components including Customs and Border Protection, Immigration and Customs Enforcement, and the Transportation Security Administration, replacing a patchwork of tools that do not share data easily.

Asha Sharma, the new Microsoft gaming CEO, has released her first internal memo, outlining her vision for the future of Xbox. She promises 'the return of Xbox' and a focus on great games, while also exploring new categories and markets. Sharma also mentions the role of AI in the future of gaming, but emphasizes that games will always be crafted by humans.