Free security scans for open-source projects
Anthropic has launched OSS Scanner, an optional service that periodically checks eligible open-source projects for vulnerabilities using its strongest models. Core maintainers can apply by submitting a pull request, and accepted projects pay nothing.
Speed comes with an important trade-off: reports are generated without human review or triage, so some may be wrong. Anthropic says each report includes a reproducer, an explanation and, when available, a proposed patch, giving maintainers material they can verify instead of a bare warning.
Run Microsoft’s coding model locally
Microsoft’s MAI-Code-1.1-Flash is available to download and run locally. Eligible coding work can stay on the user’s hardware without per-call inference charges, though Microsoft recommends more than 120GB of RAM, putting this beyond an ordinary laptop.
The model is also in production in GitHub Copilot. Microsoft says experimental access through the Copilot app, command-line interface and Visual Studio Code will arrive by the end of October.
Also
- How Notion built an internal data agent Data Scout answers plain-language questions using governed company data in Snowflake, a cloud data platform, while each employee’s own credentials determine what the agent can access.
- A clinical study of an AI intake chatbot In a study involving 98 patients, physicians monitored AMIE conversations before urgent-care visits; Google says larger clinical trials are still needed.
- Anthropic revises Claude’s usage rules Taking effect November 12, the update clarifies rules for deceptive campaigns, surveillance, weapons, high-risk decisions and AI connected to physical equipment.
Pass it on