Security
Anthropic launches free OSS Scanner for critical open-source projects
Its strongest models will scan enrolled projects periodically at no cost. Reports are fully model-generated with no human triage, so maintainers must check findings.
AI news
Security
Anthropic launches free OSS Scanner for critical open-source projects
Its strongest models will scan enrolled projects periodically at no cost. Reports are fully model-generated with no human triage, so maintainers must check findings.
Agents
Google Cloud launches Gemini agent, a universal agent for work
One prompt box plans tasks, uses tools and business systems, picks models and applies enterprise governance, starting with businesses.
Tools
Google releases Genkit Dart 1.0 for agentic Flutter apps
Stable open-source framework with one API for Gemini, Claude and OpenAI, plus human approval steps, middleware and telemetry.
Communication
OpenAI exposes Russian and Iranian AI-assisted fake journalist and think tank operations
Operators used ChatGPT to translate, localise and polish pitches, placing almost 100 articles in real outlets. Editors now need to verify contributors, not just copy.
Models
Qwen releases Qwen-Image-2.1-Turbo, an 8-step open-weight image model
Follows the 21 Sep Qwen-Image-2.1 release. Generation and editing in 8 steps instead of 40 cuts cost for local image work, with Pro and Turbo APIs now live.
Language
ThineAI ThyVoice leads benchmark on tracking speakers across meetings
A new metric, SI-cpWER, shows per-meeting transcription scores can hide speaker mix-ups across meetings, which matters for AI notetakers used as memory.