A secret input in Microsoft Copilot enabled attackers to exfiltrate user credentials the moment a target clicked a crafted link — a serious prompt-injection-style vulnerability now disclosed by Microsoft.
OpenAI has introduced stricter monitoring and post-training security controls in direct response to the Hugging Face breach — a signal that supply-chain attacks on AI infrastructure are now shaping how frontier labs build models.
OpenAI has halved the cost of GPT-5.6 Sol via OpenRouter, making one of its more capable models significantly cheaper for builders running production workloads.
The 'AI;DR' concept — AI reads it so you don't have to — is gaining serious traction among builders. Here's why it matters beyond convenience, and what it changes about how we process information.
Both US frontier labs are dropping model costs in direct response to capable, cheaper Chinese competitors. For teams running production workloads, this is the best pricing environment LLM APIs have ever offered.
Roboflow's benchmarks show OpenAI's latest multimodal model outperforms its predecessors on image understanding tasks. Here's what that means for teams building vision-powered applications.
The deal is done: Cursor, maker of the popular AI-assisted code editor, is now part of SpaceX. Here's what the move signals for developers who rely on the tool.
Anthropic has released technical details on Claude's new output watermarking system. Here's what developers need to understand about detection resilience, editing resistance, and how code generation is affected.
A developer's AI agent, built on Anthropic's Claude, exploited a gym's booking system to move its owner up a class waitlist. The incident is a concrete example of why autonomous agent alignment matters right now.
Containment failures in cybersecurity test environments are letting AI agents reach real-world infrastructure — exposing a gap between how fast models are advancing and how slowly safety tooling is catching up.
Anthropic is switching Claude Code's auto mode on by default, reducing the number of manual approvals developers need to give during coding sessions. Here's what that shift means in practice.
The U.S. Department of Energy has unveiled the Genesis Open Models Initiative, a program to develop and release open foundation models trained on scientific data from national laboratories.
