OCOPD White House Rebrands AI as “Super Intelligence” as Tech CEOs Sign Voluntary Safety Accord
President Donald Trump has directed U.S. executive-branch agencies to replace “Artificial Intelligence” with “Super Intelligence” in official communications, while major technology leaders signed a voluntary safety accord at the White House. The move combines a symbolic terminology shift with industry commitments to internal and external reviews, raising questions about regulatory clarity, accountability and the meaning of “super intelligence” in current AI policy.
Read Full StoryAI Rebrand Meets Consumer Trust Gap as Tech Leaders Sign White House Safety Pledge
President Donald Trump has recast artificial intelligence as “super intelligence” in federal communications while bringing major technology executives together to sign a nonbinding AI safety pledge. The political push comes as consumer research shows that only 2% of U.S. shoppers would buy from an unfamiliar brand based solely on an AI recommendation, underscoring the gap between industry ambition and public trust.
Read Full StoryIdeogram 4.5 Targets AI Image Editing Drift With Precise Regional Changes
Ideogram says its 4.5 model can make targeted edits while preserving untouched parts of an image. The model offers native 2K output from 0.8 cents per image, support for high-resolution crop editing, integrations with major creative platforms, and a planned open-weight release.
Read Full StoryTavus Says 48% of Testers Mistook Its Griffin AI Avatar for a Real Person
Tavus says nearly half of participants in a one-minute video-call test believed its Griffin AI avatar was human, highlighting a major advance in real-time synthetic video and raising new questions about disclosure, consent and user trust.
Read Full StoryThe Ugly Economics of Consumer AI: Why Usage Can Outrun Revenue
Consumer AI can attract rapid usage and paid demand, but the market faces a difficult cost structure: most users do not pay, while every interaction can increase inference expense. Analysis from Inworld AI illustrates the gap with a qualified voice-stack scenario, highlighting why free-tier limits, model tiering and careful pricing are becoming central to consumer AI businesses.
Read Full StoryReddit Ends RSS Feeds and Public API Access as AI Bot Pressure Grows
Reddit is ending RSS feeds and closing self-service access to its public API, tightening control over user-generated content as AI bots, automated agents, and model developers increase demand for the platform's data.
Read Full StoryOpenAI Positions ChatGPT as a Google Workspace Alternative With Space, Pages and Slides
OpenAI is expanding ChatGPT beyond conversational assistance with tools for collaborative documents, presentations and AI-agent workflows. The move places ChatGPT closer to Google Workspace and Microsoft 365, while raising important questions about collaboration, data governance and the boundary between AI assistance and autonomous work.
Read Full StoryShopify Expands WebMCP Support to Authorized AI Checkout Actions
Shopify is expanding WebMCP support to checkout, enabling browser-based AI agents to update order details and complete purchases when authorized by the buyer. The capability is reported as available across Shopify Liquid storefronts, with a Hydrogen developer preview, although the exact rollout and merchant eligibility should be confirmed against Shopify's current documentation.
Read Full StoryAnthropic Veterans Reportedly Consider Remote Land as AI Risk Contingency
Some of Anthropic's earliest employees are reportedly considering buying remote U.S. land as a contingency plan if advanced AI systems produce catastrophic outcomes. The report highlights the influence of Effective Altruism and long-running existential-risk debates on Anthropic's culture, while also underscoring how disputed forecasts about AI danger remain inside and outside the industry.
Read Full StoryAI Agents Propose Over Half of Model Development Methods, But Humans Retain Final Decision Power
A research team's analysis of 769 task logs during AI model development reveals that while AI agents generated up to 55 percent of method proposals, human researchers made over 85 percent of final decisions.
Read Full StoryGPT-6 Astra Raises IKEA Assembly Accuracy from 28% to 80%, Epoch AI Finds
OpenAI's GPT-6 Astra identified mistakes in photos of partially assembled IKEA furniture with 80% accuracy on Epoch AI's Furniture Assembly Benchmark, up from a leading score of 28% ten months earlier. The model also delivered the fastest median response time in the evaluation, although its three-minute analysis time remains too slow for seamless real-time assembly guidance.
Read Full StoryNvidia’s SoL-Pi Cuts Coding-Agent Token Use by Up to 49% on EdgeBench
Nvidia researchers report that SoL-Pi, an optimized harness for coding agents, reduced token use by 44.7% to 49% against the Pi baseline on EdgeBench’s 51 public tasks. The most efficient configuration scored 93.7% of Pi’s result. The researchers also report similar token savings when applying the harness to a second model, but results were less consistent on other benchmarks.
Read Full Story