Here are the tech and AI stories worth knowing this week, with quick summaries and links to the original reports.
LiteLLM 1.107.0 compromised in a supply chain attack
A malicious release of LiteLLM, a widely used library for routing requests across AI providers, contained a credential stealer in its compiled packages and official Docker images. The incident highlights the importance of dependency security and credential management.
Source: StepSecurity
Anthropic introduces Claude Duet, a proactive AI agent
Anthropic introduced Claude Duet as a research preview of an agent designed to connect with apps, identify tasks, and work in the background. Access is currently limited to eligible US Pro and Max subscribers through a waitlist.
Source: MacRumors
OpenAI models and Codex are coming to Google Cloud
OpenAI's models and Codex are expected to become available through Google Cloud's Model Garden and Marketplace, giving customers another way to access AI tools within their existing cloud workflows.
Source: Tech Insider
Walmart opens its store to AI agents
Walmart announced an MCP server that lets approved AI agents search its inventory and make purchases. Planned Agent Profiles will add controls such as purchase approvals and spending limits.
Source: Winzheng
Recursive releases Flint, an open-weight coding model
Recursive released Flint under the Apache 2.0 license. The model has approximately 40B total parameters, with 2.5B active at a time, and is designed to make local coding-model deployment more accessible.
Source: RuntimeWire
AI stocks fall as investors weigh infrastructure costs
AI-related stocks declined as investors assessed rising infrastructure spending and profitability concerns. The market movement highlights growing scrutiny of the costs associated with scaling AI.
Source: BNN Bloomberg
Stay in the loop
We publish a new AI and tech update every week, covering the developments that matter to founders and technical teams. Subscribe to our newsletter to get it in your inbox.