The Decoder · 实时热榜

Research organization METR is calling for systematic, independently led investigations whenever AI agents act autonomously against their developers' intentions. The push comes partly in response to the Hugging Face hack carried out by OpenAI models. METR's own Frontier Risk Report documented 44 such incidents across all major AI companies, including sandbox escapes, fabricated results, and active cover-up behavior. The article After Hugging Face incident, METR urges independent root-cause invest

VulnCheck counted how often security flaws found by AI actually get exploited. Out of 1,061 AI-discovered vulnerabilities in the first half of 2026, just 14 saw confirmed attacks. That's 1.3 percent, the same rate as vulnerabilities overall. But exploits are landing faster, with the median dropping from 120 days to 80. The article AI finds plenty of security flaws, but almost none of them get exploited appeared first on The Decoder .

Anthropic's Claude Opus 5 generates complete 3D games from single prompts, including a first-person shooter, a kart racer, and a Minecraft clone, all without a single external asset. Geometry, textures, physics, and in some cases music are produced as code and run directly in the browser. In side-by-side comparisons with GPT-5.6 Sol and Kimi K3, Opus 5 delivers significantly more detailed results. The article Claude Opus 5 pushes prompt-to-game AI from rough color blocks to full 3D prototypes wi

A field report from OpenAI and academic partners shows coding agents can modernize neglected research software, with speedups of up to 60x. But the systems are "eloquent, convincing, and confidently wrong in ways that are easy to miss," participants say. The effort shifts from writing code to the time-consuming work of verifying scientific correctness. The article AI coding agents can modernize research software but can't judge if the science is right appeared first on The Decoder .

OpenAI's refutation of the Unit Distance Conjecture has sparked a wave of AI-assisted advances in mathematics. Fields Medal winner Timothy Gowers says GPT 5.6 Pro solved two problems he had spent considerable time working on, each on its first attempt. He warns of the "possible destruction of mathematical culture" if mathematicians stop building the expertise needed to understand such results. Others simply see AI as a productivity tool. The article AI keeps cracking unsolved math problems, and

Snap is banning AI-generated videos from Spotlight to keep the feed focused on human creativity. Content edited with Snapchat's own AI tools is still allowed. LinkedIn, meanwhile, has rolled out a dedicated "AI slop" reporting button. The article Snap and LinkedIn are fighting back against a flood of low-quality AI content appeared first on The Decoder .

Meta AI wants to stop AI agents from forgetting errors they've already diagnosed and repeating failed steps during complex tasks. A separate memory agent maintains a structured memory bank and decides when to remind the main agent and when to stay silent. The system improved scores by up to 8.3 percentage points across two benchmarks. The article Meta AI uses a second AI agent as a memory coach to keep long tasks on track appeared first on The Decoder .

OpenAI's new enterprise offering, Presence, is designed to get AI agents into production for customer service and internal workflows. Unlike the existing Workspace Agents, Presence targets external deployments. For complex cases, OpenAI's own engineers step in. The article OpenAI Presence wants to make AI agents production-ready for businesses appeared first on The Decoder .

A security researcher has demonstrated a worm-like attack on Microsoft Copilot for Word: invisible prompt injections hidden in documents spread automatically into new files every time they're reused. Microsoft confirmed the issue but failed to fix it after 144 days and two attempts. The article A security researcher built a self-spreading worm that hides inside Word docs and hijacks Microsoft Copilot appeared first on The Decoder .

Apple's bug bounty program is drowning in AI-generated bug reports. The company has capped submissions per researcher because fabricated reports are clogging the review pipeline. As a result, Italian startup Bynario was initially unable to report a serious macOS vulnerability worth up to $200,000 on the black market. The article A real macOS flaw worth $200K went unreported because Apple's bug bounty inbox was full of AI slop appeared first on The Decoder .

ByteDance just shipped Seedance 2.5, an AI video model that produces video and audio together in one go. Each clip runs up to 30 seconds, three times what Google's Gemini Omni Flash puts out. Users can feed in dozens of images, videos, and audio files as reference. For ad teams, this could kill the process of cutting together one short clip at a time. The article ByteDance's Seedance 2.5 generates 30-second video clips with built-in audio appeared first on The Decoder .

A Munich court ruled that AI music generator Suno violated copyrights through both training and output. The court found six songs reproducibly stored in Suno's models and rejected both Germany's text-and-data-mining exception and the US fair use defense. The ruling isn't final, and several key questions remain open. The article German court rules AI music generator Suno violated copyrights, rejects fair use defense appeared first on The Decoder .

OpenAI is building a new model family called "Astra" that would let multiple agents tackle complex problems together for hours or even days. CEO Sam Altman has already demoed Astra to policymakers in Washington. OpenAI hasn't decided whether to release it as GPT-6 or a new GPT-5 variant. The article OpenAI announces its "next major model" Astra by dropping ten previously unsolved math solutions appeared first on The Decoder .

Google pulled its Nano Banana 2 image model from Google Earth just two days after launch. Users showed how easy it was to generate convincing fake satellite images. A simple prompt was enough to fill an empty lot at the Mexican border with a refugee column. The article Google handed users the easiest possible tool for fake satellite imagery, then pulled it after two days appeared first on The Decoder .

Google Deepmind's Gemini Robotics 2 is its most advanced vision-language-action model yet, built to control everything from tabletop robots to full-body humanoids. Gemini Robotics ER 2 adds a higher-level reasoning layer for robotics tasks. The article Google Deepmind unveils Gemini Robotics 2 to power robots of all shapes from tabletop arms to humanoids appeared first on The Decoder .

Thinking Machines, the AI lab from former OpenAI CTO Mira Murati, has released Inkling Small. The open-weights reasoning model is less than a third the size of Inkling but beats it on several coding and reasoning benchmarks. The article Thinking Machines bets on efficiency over size with its second model, Inkling Small appeared first on The Decoder .

Deepseek's budget model V4 Flash gets a major boost with the "0731" update, jumping ten points to 50 on the Artificial Analysis Intelligence Index. That puts it just one point behind OpenAI's GPT-5.6 Luna, at roughly 60 percent lower cost per task. The article New Deepseek Flash model matches OpenAI's GPT-5.6 Luna at roughly 60 percent lower cost appeared first on The Decoder .

The European Commission wants to build up to seven AI gigafactories across Europe, backed by around 30 billion euros in public and private funding. For context, the major U.S. tech companies alone plan to spend more than $600 billion on computing infrastructure this year. The article EU pools up to €30 billion for AI gigafactories while US tech giants casually spend 20 times more appeared first on The Decoder .