NerdNews

Jacob Coxon, a researcher at Anthropic, has quit his job and warned that the AI race is putting humanity at risk. He believes that the next year or two is 'crunch time' for humanity and that AI labs have a limited time to make their systems safe. Coxon recommends that OpenAI and Anthropic coordinate to limit recursive self-improvement and calls for international regulation to prevent the misuse of AI.

OpenAI, Anthropic, and Google are working together to prioritize AI safety, but their efforts are raising antitrust concerns. The companies are seeking to coordinate on safety standards, which some experts argue could violate antitrust law. The issue has sparked debate among lawmakers and regulators, with some calling for exemptions to allow for industry-wide cooperation on safety.

Nokia's Cognitive Operations platform combines edge computing and operational AI for more resilient field operations, targeting industries like mining, public safety, and defense. The platform offers real-time intelligence, predictive maintenance, and autonomous safety monitoring, and is available on Microsoft Azure Marketplace.

A researcher's resignation from Anthropic sparks concern over AI safety, with politicians and CEOs weighing in on the need for regulation and caution. The issue is gaining traction, with potential implications for the tech industry and society as a whole.

The AI risk debate has become pointlessly tribal, with the AI safety community focusing on the possibility of powerful AI models escaping human control and AI critics on the left worrying about humans using AI to exert power over other humans. However, journalist Garrison Lovely argues that these competing narratives identify essential parts of what makes AI threatening and should be considered together. Lovely suggests that the goal should be to stop the race to build technology that will make humans obsolete, and that this can be achieved by building a movement to resist replacement by AI.

Agoda replaced its 72-shard SQL Server Price Cache with DragonflyDB, an in-memory datastore, to handle growing read and write volumes and simplify scaling. The migration resulted in an eightfold improvement in P99 read latency and a less rigid scaling model.

A job applicant used ChatGPT to interview with an AI recruiter, highlighting the potential for bot-on-bot interviews in the future. The experiment showed how AI can be used to automate the interview process, but also raised questions about the effectiveness and potential biases of such systems.

A report by Mozilla found that the performance gap between open-weights models from Chinese companies and closed frontier models from US tech companies has narrowed to 4.4 months. Open models are significantly cheaper and can handle most routine work, while closed models are worth the cost for expert professional work, high-intensity retrieval, and long context tasks.

Grab's LLM-Kit framework standardizes AI agent services, reducing deployment time from 2 weeks to 1 hour. It provides scaffolding and integrations for agent development, including evaluation, tracing, and secret handling.

Apple's new Siri AI features personal search, app actions, camera questions, and saved conversations. It runs on Apple Foundation Models developed with Google and is available on iOS 27 with compatible hardware. Siri AI can find information, take actions in supported apps, and store previous conversations.

US President Donald Trump's team says AI lab leaders must implement their own industry-wide slowdown, as calls for AI regulation intensify. Anthropic CEO Dario Amodei, backed by Sam Altman and Elon Musk, requested government regulation, but the White House seems unlikely to oblige, citing concerns about ceding the lead in AI to China.

Agility Robotics' new Digit 5 robot is designed to work safely near humans without risking harm. It uses AI algorithms and multimodal sensors to detect people and can autonomously take precautions like moving to avoid them or standing still. The robot is expected to be available in 2027 and features an upgraded leg design, allowing it to lift up to 50 pound loads, and a battery that lasts 90 minutes on a single charge.

Cisco has integrated Kets Quantum Security's chip-based Quantum Key Distribution (QKD) technology into its IOS XR-based routing infrastructure, enabling scalable and quantum-secure networking for enterprises and critical infrastructure environments. The integrated service provides end-to-end encryption, increasing quantum resilience and compatibility with existing network infrastructure.

ClickFix attacks use compromised websites and fake CAPTCHA overlays to trick users into running malicious commands, infecting PCs and Macs. The technique has become mainstream, with even Kremlin-backed hacking groups using it. The attacks capitalize on user fatigue with complex web interfaces and security measures, making it easy for attackers to spread malware.

The growth of AI data centers is sparking debate across the US, with many cities opposing construction due to environmental and energy concerns. Residents in cities like Philadelphia, where big industry has left a lasting impact, are pushing back against data center plans, citing pollution, noise, and surveillance concerns. As the demand for data centers increases, cities are implementing moratoriums to assess the potential impacts, highlighting the need for sustainable and responsible development.

A mass-scanning campaign is targeting internet-exposed Vite development servers to steal cloud credentials and configurations from AWS and Azure deployments, leveraging an exploit for CVE-2026-39364. The operation has been detected by F5, with over 800 attacks and 32,000 raw events observed over a month. To protect Vite servers, developers should update to the latest version, block access through port 5173, and block suspicious requests.

Former FTC chair Lina Khan argues that existing laws can be used to prosecute CEOs of AI companies, citing a 1934 Supreme Court decision. She claims that new laws are not needed to regulate AI development, and that law enforcers already have the authority to charge companies for creating and releasing dangerous products.

Google's latest Pixel update includes features such as VIP contact widgets, expanded scam detection, and new gesture controls for the Pixel Watch. The update also brings personalized Gemini and GPS route tracking to the Pixel Watch 5. Additionally, users can customize their phones with Harry Potter Audiobook Packs.

Open Cosmos has raised €300m to expand its satellite mass-manufacturing capabilities and grow intelligence and connectivity satellite infrastructure. The company aims to provide real-time space connectivity and intelligence to governments and enterprises, enabling them to detect threats and deliver actionable insights.

Solaris' Turnstile synchronization mechanism has influenced modern system designs in web browsers and language runtimes, enabling efficient concurrency and synchronization without memory bloat.