AI Sec Watch

A mass assignment vulnerability (a type of attack where an attacker controls internal fields by sending them in a request) exists in Flowise's `/api/v1/leads` endpoint, allowing unauthenticated users to override auto-generated fields like `id`, `createdDate`, and `chatId` by including them in the request body. The vulnerability occurs because the code uses `Object.assign()` to copy all properties from user input directly into the database entity without filtering, bypassing the intended auto-generation of these fields.

London Mayor Sadiq Khan invited AI company Anthropic to expand in the city after the U.S. Pentagon designated it a supply chain risk (a label meaning the government views the company as not secure enough to work with) because Anthropic refused to give defense agencies unrestricted access to its AI tools and raised concerns about using its Claude model for mass surveillance or autonomous military targeting. The company plans to challenge the Pentagon's designation in court, and Microsoft announced it would continue using Anthropic's technology except for the U.S. Department of Defense.

Agentgateway is an open source data plane (a software layer that handles data movement for AI agents working across different frameworks) that had a security flaw in versions before 0.12.0, where user input in paths, query parameters, and headers were not properly cleaned up when converting tool requests to OpenAPI format. This lack of input validation (CWE-20, checking that data matches expected rules) could potentially be exploited, but the vulnerability has been patched.

Amazon announced that AWS customers can continue using Anthropic's Claude AI models for all work except Department of Defense projects, after the federal government labeled Anthropic a "supply chain risk." Anthropic says it will challenge this designation in court, and major cloud providers (Amazon, Microsoft, and Google) are helping customers transition to alternative AI models for defense-related work.

Google and Microsoft announced they will continue offering Anthropic's Claude AI models to their cloud customers for non-defense work, after the U.S. Defense Department designated Anthropic as a supply chain risk (a company that poses potential security or operational threats to government operations). The announcements came after the Trump administration instructed federal agencies to stop using Anthropic's technology, but the companies determined that non-defense projects are still permitted under this designation.

The U.S. Department of Defense designated Anthropic (maker of Claude AI) as a supply-chain risk after the company refused to provide unrestricted access for military applications like mass surveillance and autonomous weapons. Microsoft, Google, and AWS confirmed that Claude will remain available to non-defense customers through their platforms, and the designation only restricts direct Department of Defense use, not broader commercial applications.

The Pentagon and AI companies are in a dispute over whether existing U.S. law allows the government to use AI to analyze bulk commercial data collected from Americans for surveillance purposes. Legal experts point out that current law has a major gap: public information, commercial data (like location and browsing records), and information accidentally collected during foreign surveillance are not legally considered "surveillance," so the government can use them without warrants or court orders, even as AI makes this surveillance much more powerful than before.

Anthropic used Claude Opus 4.6 (an advanced AI model) to test Firefox's code and discovered 22 vulnerabilities, including 14 severe ones, over two weeks. Most of these bugs have already been fixed in Firefox 148 released in February, though some fixes will come in a later update. The AI was much better at finding security problems than creating working exploits to demonstrate them.

Flowise has a file upload vulnerability where the server only checks the `Content-Type` header (MIME type spoofing, pretending a file is one type when it's actually another) that users provide, instead of verifying what the file actually contains. Because the upload endpoint is whitelisted (allowed without authentication), an attacker can upload malicious files by claiming they're safe types like PDFs, leading to stored attacks or remote code execution (RCE, where attackers run commands on the server).