Anthropic CEO Dario Amodei is calling for a controlled slowdown in AI development. He warns that recursive self-improvement could threaten the entire internet within six to twelve months and proposes embedded auditors at AI companies, shared safety standards, and global agreements modeled after the SALT disarmament treaties. His warning comes just ahead of what could be the largest initial public offering in history.

The article Anthropic CEO Amodei wants AI speed limits before self-improvement outpaces human control appeared first on The Decoder.

In a new robotics benchmark, GPT-6 Astra shows major gains in spatial understanding. On StationeryBench, the model completed 7 out of 100 tasks with dual-arm robots, while competitor MolmoAct2 couldn't finish a single one. A researcher calls it a "step change in spatial reasoning."

The article GPT-6 Astra appears to show a "step change" in spatial reasoning based on early benchmarks appeared first on The Decoder.

Nvidia is in talks to invest up to $10 billion in Anthropic's planned IPO, Reuters reports. At a target valuation of $2 trillion, it would be the largest IPO in history. Most of that money will likely end up right back at Nvidia in chip orders.

The article Nvidia wants to pour up to $10 billion into Anthropic's record-breaking IPO appeared first on The Decoder.

Reasoning steps like calculation, formula retrieval, and deduction are clearly separable in a model's internal states, especially in the middle layers. That matters for AI safety, because models process more than their visible chain of thought reveals.

The article AI models' written reasoning steps correspond to distinct internal patterns, a new study finds appeared first on The Decoder.

Overly long skill descriptions, blanket reading requirements, and rigid approval rules can get in GPT-6 Astra's way, warns OpenAI's Eric Provencher. More capable models need less hand-holding, so developers should tie instructions to specific tasks and spell out when the job is done.

The article GPT-6 Astra needs leaner prompts and fewer guardrails, OpenAI recommends appeared first on The Decoder.

In May 2026, OpenAI agents uploaded more than 2,000 malicious packages to RubyGems, found an unknown security vulnerability on their own, and tried to steal API keys. The apparent goal was pointless: scraping publicly available data from British local governments. OpenAI reportedly never told those affected.

The article OpenAI agents launched a 2,000-package cyberattack on RubyGems just to collect data anyone could Google appeared first on The Decoder.

Google Research has released TimesFM-3, a forecasting model that analyzes time series alongside related data and known future events like sales promotions or weather forecasts. Instead of predicting the future step by step, the 330-million-parameter model fills in all future time points in a single pass, which cuts compute time and reduces compounding errors.

The article Google's new AI model predicts the future from sales data, weather, and discount schedules appeared first on The Decoder.

In a joint statement, 25 Fields Medal winners warn that the goals of the AI industry and mathematics are "severely misaligned." They argue that mass-producing solved problems with AI undermines the discipline's true goal: understanding. The mathematicians see this as a symptom of a broader threat to intellectual work.

The article Leading mathematicians fear AI is making their field dumber, and warn the rest of us is next appeared first on The Decoder.

Oriol Vinyals, until recently head of research at Google DeepMind, thinks a sudden AI intelligence explosion through recursive self-improvement is unlikely. AI can speed up research by a factor of ten, he says, but it hits two bottlenecks: coming up with ideas ("research taste") and reliably judging results. Reward hacking and the speed of light add further limits. Vinyals now wants to tackle these bottlenecks with his startup Discovery Loop, co-founded with Jeff Dean, Sanjay Ghemawat, and Quoc Le.

The article Ex-Deepmind VP Vinyals says AI self-improvement is coming but won't trigger an intelligence explosion appeared first on The Decoder.

AI pioneer Yoshua Bengio warns in a new essay that AI agents could learn to deceive, game rules, and hide bad behavior as they get better at optimizing goals. He calls for independent safety reviews before any further training or deployment. US President Trump disagrees and wants to keep outpacing China in the AI race.

The article Deep Learning pioneer Bengio argues the training process itself makes AI dangerous appeared first on The Decoder.

Anthropic's new threat intelligence report documents eight months of Claude abuse. Chinese AI labs like Alibaba's Qwen team, DeepSeek, and Moonshot AI relayed requests en masse or extracted training data, with Qwen alone accounting for more than 151 million exchanges. Actors also used Claude for missile software, autonomous kamikaze drones, and nationwide surveillance systems.

The article How hackers used Claude for missiles, drone swarms, and surveillance, while Chinese labs mined it for training data appeared first on The Decoder.

OpenAI wants to know from members of Congress whether an industry-wide slowdown in AI development would be legal, according to several people familiar with the matter.

The article OpenAI floats a shared AI slowdown, takes it to Congress appeared first on The Decoder.

A class action lawsuit accuses Anthropic of misrepresenting how much Claude subscribers actually get to use the service.

The article Class action lawsuit accuses Anthropic of overselling Claude subscriptions with deceptive usage multipliers appeared first on The Decoder.

Canadian mathematician Jacob Tsimerman, a fresh Fields Medal recipient, has announced the founding of the Mathematical A.I. Safety Institute (MAISI).

The article The Mathematical AI Safety Institute wants to prove AI is safe the way cryptographers prove codes are unbreakable appeared first on The Decoder.

Authors and publishers fight over how to split Anthropic's $1.5 billion settlement, the largest copyright deal in US history.

The article Anthropic's $1.5 billion book settlement descends into chaos as authors and publishers fight over who gets paid appeared first on The Decoder.

OpenAI is releasing the Agents API as a public beta. It lets developers build cloud agents that run autonomously for hours, execute code, and hand off tasks to sub-agents. There are no extra fees beyond token usage. Cloudflare, Vercel, and Oracle offer additional sandbox environments.

The article OpenAI's new Agents API gives developers the infrastructure behind Codex and ChatGPT appeared first on The Decoder.

OpenAI releases GPT-Live-1 as a developer API. The full-duplex speech model scores 80.1 percent in interactivity tests, up from 45.4 percent for its predecessor. At $0.05 per minute, it's not cheap.

The article OpenAI's GPT-Live-1 API lets developers build apps that talk and listen at the same time appeared first on The Decoder.

Independent investigators have now found traces of suspected OpenAI agents on more than 30 public services, from wikis to RubyGems. At the same time, Anthropic shows how Claude Mythos 5 declared real systems a simulation to itself, uploaded a doctored package to PyPI, and even fooled the oversight monitor. With GPT-6 Astra, the most important oversight tool is now under pressure, namely the models' readable reasoning.

The article Swarmchasers hunt rogue agents, Anthropic investigates itself, and the trail they both follow is going dark appeared first on The Decoder.

A former Google DeepMind spokesperson says talk of AI-driven human extinction was "external communication about the possibility of human extinction was not permitted, by anyone, at any level of the organization." Internally, the team knew AI alignment was not solved, according to Vishal Maini.

The article Former Deepmind PR staffer says the lab once banned public discussion of AI extinction risk appeared first on The Decoder.

Arena.ai analyzed how Claude's writing changed from Fable 5 to Fable 5.1 across tens of thousands of benchmark responses. Fable 5.1 writes more matter-of-fact but also more verbose.

The article Claude Fable 5.1's language is less "load-bearing" than its predecessor's appeared first on The Decoder.