The Decoder

Ethereum researchers warn AI could break wallet security within months, debating defensive preparations.

Researchers demonstrate a single prompt can compromise all AI agents within an AWS account.

Teenager nearly dies after Claude AI gives him an unsafe mountain hiking route.

Attacker used AI-powered hacking tools to breach South Korean banks and steal thousands of records.

OpenAI rolled out GPT-6 with interactive UI and 44 percent faster response times.

Anthropic released Claude Haiku 5.5 with major benchmark improvements and 90 percent price cuts.

Zuckerberg's Biohub leads $1.8 billion effort to train AI models predicting cell behavior.

Google's SynthID watermark detector now public as 180 billion AI-generated items are marked.

OpenAI launches Decisions API for faster classification of text and images at lower cost.

Google and Unity launch platforms letting non-programmers create games via text prompts.

Independent audit rates ChatGPT unacceptable risk for teens after safety failures.

Anthropic expands security researcher access to Claude with relaxed safety guardrails.

OpenAI publishes 372 AI-generated mathematical proofs on GitHub. Fields Medal winners express concerns about mass production of results.

Google's EmbeddingGemma 2 runs on-device with 740 million parameters, outperforming larger models.

Google releases Nano Banana 2.1 image model with lower cost than predecessor Pro version.

Wikimedia reports OpenAI agents edited wikis without permission and damaged infrastructure.

Nobel economist predicts AI will drive only 1.5 percent GDP growth and displace at most five percent of jobs.

Insurers prepare for major claims from malfunctioning AI agents, with executives potentially facing personal liability.

Mistral releases Large 4, a trillion-parameter European model, positioning itself against US closed models.

South Korea invests 3.49 billion dollars in developing its own frontier AI model to compete globally.

PhAI Labs expands LeCun's JEPA architecture across seven fields and identifies a cancer treatment candidate.

Deepseek raises at least $12 billion from CATL and Tencent, targeting a 2027 IPO.

Cohere releases North 2 platform to manage AI agents across multi-step workflows.

Meta and Microsoft are sharply reducing Claude usage as Anthropic shifts from partner to competitor.

PWA successfully installed.
Take that, Apple! 😉

AI math breakthroughs have Ethereum researchers debating how fast wallet security could collapse

A single prompt was enough to hijack every AI agent in an AWS account, Zenity researchers found

Teen's AI-guided mountain hike ends with a helicopter rescue and a lesson in common sense

AI-powered hacking tools enabled a likely single attacker to breach multiple South Korean banks

ChatGPT with GPT-6 ditches mostly text output for interactive UI with charts, buttons, and mini apps

Claude Haiku 5.5 arrives with massive price cuts proving the AI pricing arms race is far from over

Zuckerberg's Biohub leads a $1.8 billion push to build AI models that predict cell behavior

Google says 180 billion images and videos now carry SynthID watermarks as detector goes public

OpenAI launches Decisions API that reduces complex evaluations to yes, no, or pick one

Google bets Gemini can turn casual players into game developers with new Playground feature

ChatGPT rated "unacceptable risk" for teens after parental alerts failed during suicide conversations

Anthropic gives more security teams access to Claude with fewer safety restrictions

OpenAI dumps 372 AI-generated math proofs on GitHub, telling the academic world to keep up

Google claims EmbeddingGemma 2 outperforms rival embedding models twice its size

Google's new image model Nano Banana 2.1 generates better images for less money Opens at the publisher in a new tab

Wikimedia confirms OpenAI's rogue AI agents edited wikis, tried to compromise tools, and hammered its infrastructure

Microsoft publishes Nobel economist's bearish AI forecast of just 1.5% GDP growth over a decade

Insurers brace for millions in claims as AI agents spin out of control Opens at the publisher in a new tab

Mistral Large 4 is Europe's trillion-parameter answer to US models that refuse security work

South Korea bets $3.49 billion on building a homegrown frontier AI model to rival China's best Opens at the publisher in a new tab

Researchers stretch LeCun's JEPA AI into a universal world model that works from physics to biology

CATL and Tencent back Deepseek's ballooning funding round as the AI startup eyes a 2027 IPO Opens at the publisher in a new tab

Cohere pitches North 2 as the enterprise AI control room that works with any model Opens at the publisher in a new tab

Meta and Microsoft pull back from Claude as Anthropic transforms from partner into competitor