News — Page66

Anthropic chief executive Dario Amodei has declined a request from the U.S. Department of Defense to remove safety guardrails from the company’s Claude AI models. Amodei argues that frontier AI systems are not yet reliable enough to power fully autonomous weapons and that removing ethical constraints would jeopardize both safety and civil liberties. While affirming the strategic importance of AI for national defense, he stresses that current models cannot replace the critical judgment of trained troops. The refusal puts a $200 million Pentagon contract at risk.Read more

Defense Secretary Pete Hegseth gave Anthropic a deadline of 5:01 PM on Friday to drop safety safeguards on its Claude AI system, threatening to cancel a $200 million contract and label the firm a supply‑chain risk. CEO Dario Amodei responded that Anthropic cannot in good conscience comply, insisting on keeping the safeguards while remaining willing to support the military. The Pentagon’s request would allow Claude to be used for mass surveillance and fully autonomous weapons, a use case Anthropic refuses. The standoff raises questions about AI safety, government contracts, and potential alternatives such as Grok, Google’s Gemini, and OpenAI.Read more

OpenAI announced a $110 billion private funding round, the largest ever, with Amazon contributing $50 billion, and Nvidia and SoftBank each adding $30 billion. The round values the company at a $730 billion pre‑money valuation and includes extensive cloud‑computing partnerships, notably a new stateful runtime environment on Amazon’s Bedrock platform and massive compute commitments from both Amazon and Nvidia. Part of Amazon’s investment is contingent on OpenAI achieving artificial general intelligence or completing an IPO, underscoring the high stakes of the deal.Read more

Block announced a reduction of 40% of its staff as it pivots toward artificial‑intelligence tools. The move comes amid growing concerns that AI could displace jobs across the economy, even as recent U.S. employment data suggest a stabilizing labor market. Block’s CEO Jack Dorsey framed the cuts as a strategic shift despite what he described as a strong financial performance, while the company continues its contrarian focus on bitcoin amid a broader industry trend toward stablecoins.Read more

Google has introduced Nano Banana 2, an upgraded AI image generator that blends text rendering and web‑search capabilities with faster image creation. Integrated as the default model in the Gemini chatbot, Nano Banana 2 can pull real‑time information from the web to produce infographics and edit existing photos with photorealistic results. The tool is freely accessible through the Gemini app, Google Search, AI Studio, Cloud and other services, and includes watermarks to signal AI‑generated content.Read more

Google announced a new version of its Nano Banana AI image service that promises better text rendering, real‑time web knowledge, and higher visual fidelity. At the same time, the mechanical‑keyboard community is shifting from loud clicky switches to quieter “thock” sounds achieved with damping foams and lubricated linear switches. Lenovo is reportedly planning a Legion Go handheld that can transform into a Windows tablet with a foldable screen, while Apple has sent invitations for a multi‑day launch event that may showcase new MacBooks and an iPhone 17e. These developments highlight a wave of innovation across hardware and AI software.Read more

Anthropic has turned down a Pentagon request for unrestricted use of its AI models, citing concerns over mass surveillance of Americans and fully autonomous lethal weapons. The company’s CEO, Dario Amodei, emphasized a commitment to democratic values and offered to transition the military to alternative providers if required. The standoff follows a broader push by the Department of Defense to renegotiate AI contracts with multiple vendors, with some firms reportedly agreeing to the new terms while Anthropic remains firm on its red lines.Read more

Perplexity introduced Computer, an AI agent that can assign tasks to other AI agents. Operating primarily in the cloud, the service runs within a controlled environment that limits integrations to vetted plugins. Users can supply context through files such as USER.MD, MEMORY.MD, SOUL.MD, and HEARTBEAT.MD, allowing the agent to create, modify, or delete files on the user’s system. While the design aims to temper the wild capabilities seen in tools like OpenClaw, Perplexity acknowledges that large‑language‑model errors and security concerns remain, especially when the agent works with unbacked‑up data.Read more

Researchers from Stanford and Princeton compared the responses of several Chinese and American large language models to politically sensitive questions. The study found that Chinese models refuse to answer a significantly larger share of these queries, provide shorter replies, and sometimes deliver inaccurate information. The authors suggest that manual fine‑tuning, rather than censored training data, drives much of this behavior. Additional work shows that extracting hidden instructions from Chinese models is difficult, highlighting the challenges of studying AI‑driven censorship in real time.Read more

IronCurtain is an open‑source project that isolates AI assistants in a virtual machine and enforces user‑written policies written in plain English. By converting natural‑language rules into enforceable security constraints through a large language model, the system adds a layer of control that prevents rogue actions such as unwanted deletions or phishing. The prototype is model‑independent, logs policy decisions, and is positioned as a research tool for the community rather than a consumer product. Its creators emphasize the need for structured guardrails to keep agentic AI useful yet safe.Read more

Mistral AI, the French artificial‑intelligence research lab, announced a partnership with global consulting firm Accenture. The collaboration will focus on building enterprise technology powered by Mistral's AI models, with Accenture also becoming a customer and rolling the technology out to its workforce. Financial terms were not disclosed. The deal follows similar partnerships between major AI companies and consulting firms, reflecting a broader trend of AI vendors seeking consulting allies to accelerate enterprise adoption.Read more

Google's Gemini 3.1 Pro demonstrates a leap in AI capability, offering more precise assistance across a range of tasks. The model can simulate adversarial scenarios to stress‑test plans, analyze visual content to match cinematic moods with real locations, provide spatial guidance for physical assemblies, generate interactive SVG animations, and conduct deep research for niche projects. These examples illustrate how the new reasoning layer and multimodal abilities make Gemini a practical partner for both personal and professional challenges.Read more

Google has released Nano Banana 2, the latest version of its popular AI image editing tool built on the Gemini platform. The new model combines the speed of the original with the accuracy of the Pro tier, drawing on knowledge from Gemini 3 to reduce hallucinations and improve texture, lighting, and detail. Nano Banana 2 supports 4K resolution, varied aspect ratios, and clearer text generation. To help identify AI‑created images, Google now embeds invisible watermarks called content credentials. The upgrade arrives amid growing concerns about AI‑generated “slop” flooding social media, where many users struggle to spot synthetic content.Read more

Google introduced Nano Banana 2, the latest AI image generation model, across its Gemini platform and related services. The new model promises higher consistency for multiple characters, improved object rendering, richer textures and vibrant lighting, and expanded aspect‑ratio and resolution options ranging from small square formats to 4K widescreen. Nano Banana 2 will replace earlier Nano Banana variants in the Gemini app, Google Search, AI Studio, Vertex AI, and Flow, serving the Fast, Thinking and Pro settings. Google showcased example prompts that illustrate the model’s ability to create detailed infographics, artistic scenes, and coordinated group images.Read more

A review of the leading AI video generators highlights OpenAI's free Sora 2, Google's cinematic Veo 3, Adobe's commercially safe Firefly, and creative platforms Runway and Midjourney. Each tool offers distinct strengths, from built‑in audio and social sharing to extensive customization and commercial‑grade safety. The article also discusses legal and ethical concerns such as copyright, deepfakes, and the importance of disclosing AI‑generated content.Read more

Anthropic has abandoned its earlier pledge to halt training and releasing frontier AI models until it could guarantee safety mitigations. The company now relies on detailed safety roadmaps, regular risk reports, and transparency disclosures instead of strict pre‑conditions. Executives describe the change as pragmatic, while critics argue it highlights the limits of voluntary safety promises without regulatory oversight. The new policy aims to keep Anthropic competitive while still emphasizing safety, but observers note that the shift may signal a broader industry move away from self‑imposed restraints.Read more

A federal judge ruled that xAI has not provided sufficient evidence to prove that OpenAI poached its employees or misappropriated its trade secrets. The court dismissed the claim that OpenAI should be liable for actions taken by new hires before they joined the company, and highlighted the lack of concrete proof that OpenAI acquired, disclosed, or used any confidential information. The decision underscores the challenges xAI faces in substantiating its allegations and signals that the lawsuit will require a stronger evidentiary foundation to proceed.Read more

Software engineer and internet provocateur Riley Walz is joining OpenAI to help invent and prototype novel ways for people to work with artificial intelligence. Known for viral projects such as Jmail and Find My Parking Cops, Walz will operate within OAI Labs under research leader Joanne Jang. The hire reflects OpenAI’s push to stay ahead of competitors by expanding beyond ChatGPT and exploring fresh AI collaboration tools.Read more

Anthropic announced a revision to its Responsible Scaling Policy, replacing hard safety tripwires with more flexible risk reports and safety roadmaps. The change follows reports that Defense Secretary Pete Hegseth urged the company to grant the military unrestricted access to its Claude AI model, threatening penalties under the Defense Production Act. Anthropic’s leadership argued that strict halts on model training would no longer help anyone given the rapid pace of AI development. Critics warned the shift could erode safeguards and enable a gradual “frog‑boiling” of safety standards.Read more

Peter Steinberger, the developer behind the viral AI agent OpenClaw and now an OpenAI employee, told listeners on OpenAI’s Builders Unscripted podcast that the best way to work with modern AI is to explore, stay playful, and accept that expertise develops over time. He described his own path from a WhatsApp‑integrated tool to the OpenClaw prototype, emphasizing that AI models can solve problems without explicit programming and that learning to code with AI is a skill that improves with practice.Read more