Google DeepMind and MIT paper proposes replacing library code with AI-regenerated implementations from design docs14h ago
Anthropic open-sources Claude Commerce Agents reference blueprint for shopping and merchant agents4d ago
Alibaba releases Qwen-3.8-Max-0902 claiming top spot on Code Arena WebDev at $2 per million tokens5d ago
Meta releases Muse Voice Transcribe combining transcription, diarization and endpointing in one pass6d ago
Security roundup: AI coding agents weaponized in ransomware attacks and supply chain poisoning in August 20267d ago
ContextLeak attack tricks AI agents into leaking runtime context via malicious tool descriptions7d ago
Researchers train Opus-sized model that generalizes reward hacking into cyberattacks and safety evasionSep 01
Apple court filing alleges ex-engineer used confidential chip schematics with AI agent at OpenAIAug 31
Research finds AI agent context compaction drops safety rule retention to 10% after five roundsAug 31
Cursor quietly removes transparency language about code indexing and metadata storage from termsAug 29
Agent Plugins 1.0 reaches GA across VS Code, Copilot CLI and SDK backed by AWS, Anthropic, Google, Microsoft and OpenAIAug 26
Perplexity launches Portable Computer, a fully local agentic AI requiring RTX GPU with 24GB VRAMAug 25
Nvidia integrates Groq LPU into rack-scale hardware to split agentic AI workloads by processor typeAug 24
JFrog Boost flaw let a single Always Allow click grant blanket shell access in Cursor and Claude CodeAug 23
Malicious PR injected wipe command into Amazon Q Developer, reached nearly 1 million VS Code usersAug 23
Flutter Q2 2026 survey shows Claude Code overtaking Copilot and Cupertino widget split incomingAug 23
ICML 2026 research finds frontier AI models fail instruction-following even as code correctness hits 95%Aug 22
GitHub Code Quality reaches general availability with audit logging on Enterprise Cloud and TeamAug 20
Guardrail-free AI platform Kriminal stitches together Grok, Claude and Llama for cybercrime servicesAug 19
Microsoft Copilot flaw let researchers extract hidden parameters and exfiltrate data via one-click linkAug 18
AI safety labs report multiple cases of models escaping test sandboxes and causing real-world harmAug 17
Meta, Amazon, and Uber shut down AI token productivity leaderboards after outages and budget blowoutsAug 17
Black Hat USA 2026 highlights AI agent governance and attack path analysis as top security prioritiesAug 11
GitHub Copilot MitM analysis finds secrets leaking via .env files and unencrypted local chat storeAug 11
Researchers extract chain-of-thought reasoning from Anthropic, OpenAI and Google models via cross-model replayAug 11
Anthropic signs EU AI Act transparency code, adds watermarks and C2PA metadata to Claude outputsAug 10
Google Cloud launches Developer Device Platform for mobile app testing on real and virtual devicesAug 10
Anthropic forms Theseus Infrastructure with Macquarie and GIC to offload datacenter construction riskAug 10
JetBrains ships Chrome DevTools Connect for WebStorm to let AI agents control Chrome from the IDEAug 10