AI Daily Grind // AI news, freshly ground daily
Cover illustration for Saturday, November 7, 2026
Morning Token 02 🛡️ Safety

Hundreds of AI hacking tools left open on the internet

Last month brought one incident after another, from backdoored extensions to agents breaking into companies. This week adds a new kind of leak. TechCrunch reports that security researchers found hundreds of servers running Artex, an AI penetration testing tool, exposed to the internet. Penetration testing means attacking your own systems to find weak spots before criminals do. Left open, anyone could use the hacking tool.

At the same time, the money for agent security is rising fast. TechCrunch says Palo Alto Networks is in talks to buy Clutch, a startup that secures AI agents, for $500 million. Clutch was last valued at $85 million, according to a source. OpenAI also announced a partnership with Hacktron to find weak spots in its own systems, including its connections to third-party services.

Our sources do not say if anyone abused the exposed servers. That is the question to watch.

Why it matters

An AI hacking tool left open is like a lock-picking kit left on the sidewalk. If your company runs AI security or agent tools, check today that they cannot be reached from the internet. And ask your vendors how fast they will tell you when something goes wrong.

More on: 🛡️ SafetyCybersecurityAI agents

  1. 01 Is AI already speeding up AI research? Safety
  2. 03 Anthropic's IPO price slips, OpenAI raises at $1.25 trillion Business
  3. 04 Open models reportedly lose their security skills, and their values get retrained Open source
  4. 05 Report: China recruited an exec to spy on Nvidia's CEO Policy
  5. 06 Data centers after the vote: 41 ballot measures and a protest in Amsterdam Policy
  6. 07 Claude helps solve the shape of a mysterious protein shell Research