News · Daily briefs

AI security news, with the fix.

Each brief covers one real AI security story: what happened, who is affected, why it matters and what to do now. Every brief links to the original source and credits the researchers who found the problem.

Real incidents onlySources linked in every briefNew stories every day
In plain words

Short, sourced briefs on new AI security vulnerabilities and incidents, from prompt injection and AI agent flaws to MCP server bugs, AI supply chain attacks and deepfake fraud. Each brief says what happened, who is affected, why it matters and what to do now, and links to the original advisory or report.

Stories
11
Topics
LLM apps · AI agents · MCP · Voice AI
Sources
Linked in every brief
Feed
RSS at /news/feed.xml
Latest briefs

New AI flaws, newest first.

Follow along on LinkedIn, or subscribe to the RSS feed at hanulsec.com/news/feed.xml.

  1. 3 min readAI security newsAn AI server-admin tool could run any command on your servers.mcp-ssh-manager, an MCP server that lets AI assistants manage servers over SSH, built backup commands from unescaped input. Whoever controls the agent’s tool arguments could run their own commands on every managed server.
  2. 3 min readAI security newsAn AI document reader could reach files outside its folder.AWS disclosed a path traversal in awslabs.document-loader-mcp-server, the MCP server that lets AI assistants read PDF, Office and image files. In some setups its folder restriction was not enforced.
  3. 3 min readAI security newsAn AI gateway flaw is now on CISA’s exploited list.On 18 November 2026, CISA added a LiteLLM sandbox escape to its Known Exploited Vulnerabilities catalogue. If your company routes AI traffic through LiteLLM, this is a patch-now issue.
  4. 3 min readAI security newsA malicious MCP server could collect your secrets.The official Model Context Protocol (MCP) Python SDK had an OAuth flaw that let an untrusted MCP server receive credentials meant for someone else. Updating fixes most of it, but not all.
  5. 3 min readAI security newsA research paper hid orders for AI reviewers.Hidden text in a published research paper told AI reviewers to write a positive review and avoid negatives. The same trick works on any AI that reads documents written by outsiders.
  6. 3 min readAI security newsAn AI agent turned a security patch into an exploit on its own.Google’s Threat Intelligence Group found an attacker’s framework that compared software before and after security updates, wrote working exploits and ran them with no human in the loop. Google calls it a first.
  7. 3 min readAI security newsA “read-only” AI tool could write files anywhere.AWS fixed a path traversal in its Pricing MCP Server’s cost report tool. The tool was marked read-only, which can lead AI clients to run it without asking the user.
  8. 3 min readAI security newsAn MCP gateway could be steered into your internal network.A server-side request forgery flaw in mcp-gateway-registry, an open-source MCP gateway, let a user who could register servers make the gateway call internal services.
  9. 3 min readAI security newsA breach at an AI vendor exposed customers’ chats.Notion says AI chat transcripts from its Insights feature were exposed after attackers reached AI provider Fireworks AI through a third-party cloud provider.
  10. 3 min readAI security newsCloned executive voices, real wire transfers.US prosecutors charged three men over an alleged scheme that used AI to clone the voices of executives and lawyers to get finance staff to wire money.
  11. 3 min readAI security newsA security setting is only as strong as the app around it.Check Point researchers found a sandbox escape in ChatGPT’s HTML previews that got around Lockdown Mode, the setting for the most sensitive users. OpenAI fixed it after responsible disclosure.
Questions

About these briefs.

How do you choose the stories?
We pick new, verified AI security vulnerabilities and incidents that affect companies building or using AI: chatbots, AI agents, MCP servers, AI gateways, model supply chains and voice AI. We only cover real events with a public source.
Who writes the briefs?
The HanulSec AI red team. Each brief credits the researchers who found the issue and links to the original advisory or report, so you can check the details yourself.
Can I get these by RSS?
Yes. The feed is at https://hanulsec.com/news/feed.xml and updates with every new brief.
Request a quote

Scope it in two minutes.

Tell us what you need protected. A security engineer, not a salesperson, replies with a fixed-price proposal.

  • Reply within one business day
  • Free 30-minute scoping call
  • Mutual NDA before any detail is shared
  • Fixed price, free retest of fixed findings