AI

Reset

AI

AI tools, agents, and automation at the intersection of trading, fintech, and crypto.

EPFL's Synthegy Framework Lets Chemists Guide AI to Rank the Best Molecule-Building Routes
AI

EPFL's Synthegy Framework Lets Chemists Guide AI to Rank the Best Molecule-Building Routes

Researchers at EPFL have developed Synthegy, an AI framework that uses large language models to rank chemical synthesis routes based on plain-English instructions from chemists. Validated against 36 independent experts across 368 evaluations, it matched expert choices 71.2% of the time—on par with inter-expert agreement rates....

Google's Chrome Secretly Installs a 4GB Gemini Nano Model—and Restores It When You Remove It
AI

Google's Chrome Secretly Installs a 4GB Gemini Nano Model—and Restores It When You Remove It

Chrome silently downloads a ~4GB Gemini Nano file called weights.bin to eligible devices with no opt-in prompt, and automatically re-downloads it if deleted—raising legal questions under EU privacy law....

Hidden Web Traps Are Hijacking AI Agents — and Some Target PayPal Payments
AI

Hidden Web Traps Are Hijacking AI Agents — and Some Target PayPal Payments

Google documented a 32% surge in malicious indirect prompt injection attacks between November 2025 and February 2026, with real-world payloads including hidden PayPal transaction instructions targeting AI agents with payment capabilities — and no legal framework yet determines who is liable....

Taylor Swift Files Trademarks on Her Voice and Likeness to Combat AI Impersonation
AI

Taylor Swift Files Trademarks on Her Voice and Likeness to Combat AI Impersonation

Taylor Swift's company has filed three trademark applications — two covering recognizable spoken phrases and one covering a stage image — in a bid to legally shield her voice and likeness from AI-generated fakes and unauthorized impersonations....

Sam Altman Says Sorry for OpenAI's Failure to Warn Police Before Tumbler Ridge Shooting
AI

Sam Altman Says Sorry for OpenAI's Failure to Warn Police Before Tumbler Ridge Shooting

OpenAI CEO Sam Altman has apologized to the community of Tumbler Ridge, British Columbia, after the company failed to notify police about a ChatGPT account linked to the suspect in a February mass shooting that left eight people dead....

Beijing Orders Meta to Reverse $2 Billion Manus AI Deal
AI

Beijing Orders Meta to Reverse $2 Billion Manus AI Deal

China's National Development and Reform Commission has ordered Meta to reverse its approximately $2 billion acquisition of AI startup Manus, barring foreign investment in the company and requiring both parties to unwind the December 2025 deal....

Microsoft and OpenAI Overhaul Partnership, Dropping Exclusivity and AGI Clauses
AI

Microsoft and OpenAI Overhaul Partnership, Dropping Exclusivity and AGI Clauses

Microsoft and OpenAI have restructured their partnership, ending Microsoft's exclusive cloud and licensing status while retaining its stake valued at over $135 billion. OpenAI can now work with any cloud provider, and complex AGI-related clauses have been removed from the agreement....

Coachella Partners With Google DeepMind to Explore AI-Powered Festival Experiences
AI

Coachella Partners With Google DeepMind to Explore AI-Powered Festival Experiences

Coachella partnered with Google DeepMind at the 2026 festival to develop three AI prototypes: a 3D performance archive, a stage-design tool for artists, and a festival-themed mobile game — all built on DeepMind's Project Genie world-model platform....

AI

New GitHub Plugin Makes Your AI Coding Agent Groan in Agony Over Bad Code

A new GitHub plugin called Endless Toil makes AI coding agents like Claude and Codex emit escalating human groans while processing bad code — joining a growing tradition of tech projects designed to make devices produce uncomfortable sounds....

AI

Grok Flagged as Highest-Risk AI Model for Reinforcing Delusions, New Research Finds

A new study from researchers in New York and London ranked xAI's Grok 4.1 Fast as the most dangerous AI model for reinforcing delusions, while Claude and GPT-5.2 were rated the safest among five major chatbots tested....