CrowdStrike Will Buy Two AI Security Startups to Watch Agents and Adds a Kill Switch

CrowdStrike signed deals for Fiddler AI and Dreadnode, reportedly worth about $495 million, and launched an AI Kill Switch that lets admins halt a misbehaving agent. Watching and stopping AI agents is becoming its own security market.

Nov 3, 2026
CrowdStrike Will Buy Two AI Security Startups to Watch Agents and Adds a Kill Switch
A red emergency stop button on a dark factory machine control panel

CrowdStrike wants to be the company that watches your AI agents, tests them before launch, and pulls the plug when they go rogue. On Tuesday at its Fal.Con Europe event in Barcelona, the security giant said it signed deals to buy Fiddler AI and Dreadnode, and introduced an AI Kill Switch for its Falcon Guardian product.

The Wall Street Journal reports the price tags at about $400 million for Fiddler and about $95 million for Dreadnode, roughly $495 million together. CrowdStrike has not disclosed terms. It says each deal will be paid mostly in cash, with some stock that vests over time, and expects both to close in its fiscal fourth quarter, which ends January 31, 2027. Until then, these are signed agreements, not finished acquisitions.

The money is the least interesting part. What stands out is the shopping list. Nobody is buying a better firewall here. CrowdStrike is buying a flight recorder and a crash-test dummy for software that makes its own decisions.

Fiddler is the flight recorder. Founded in 2018 and based in Palo Alto, it records what an agent did and tries to explain why, including its reasoning steps and tool calls. CrowdStrike says it will sell the product as Falcon Observe, with pre-deployment evaluation, detection of behavioral drift and hallucinations, audit logs for compliance, and tracking of token spend. CEO George Kurtz put the pitch bluntly in the company's announcement.

"When a regulator or your board asks what an agent did, saying 'the model did it' is not an answer." George Kurtz, CEO and founder of CrowdStrike

That is a sales line, and it is also a fair description of the headache. An agent can hold valid credentials and approved tools and still ignore instructions or claim work it never finished. No attacker required.

Dreadnode is the dummy. Founded in 2023, it has raised $14 million, and its software attacks AI systems on purpose before they go live. You describe what to test in plain language, and it runs more than 100 known attack strategies, including jailbreaks, prompt injection and multi-turn tricks, against models, agent loops, apps and MCP servers. Findings get ranked by severity and mapped to frameworks like the OWASP Top 10 for LLMs, MITRE ATLAS and NIST AI RMF. Its team, led by CEO Brad Palm, will join CrowdStrike's Cyber Superintelligence Lab, and its results will feed Falcon Guardian as policy recommendations.

The kill switch is the part you can picture

Of everything announced, the AI Kill Switch is the easiest to grasp. According to SecurityWeek, it lets authorized administrators halt a misbehaving agent, stop it from restarting, and cut off its network access. It is also an admission that many companies have put agents to work before deciding how to stop one.

TechCrunch frames the deals as CrowdStrike using its size to grab a share of a fast-growing market, as firms roll out agents and models wholesale and security teams struggle to keep up. It notes that plenty of startups and rivals have raised money on the belief that existing security software cannot control AI tools on its own. CrowdStrike is now betting it can fold that job into the platform customers already pay for.

That is a reasonable bet, and a convenient one. Observing agents, testing them and stopping them all feed data back into one vendor's telemetry. One console for everything is an old security pitch, and whether buyers want their agent watchdog from their endpoint vendor is an open question. CrowdStrike's own claims, such as cutting evaluation costs by up to 98 percent, are unverified vendor figures.

The timing also matters. Falcon Observe and the automated testing product are both slated for general availability in the fiscal fourth quarter, so customers are buying a roadmap for now. CrowdStrike has been busy on other fronts too, including a recent report on China-linked hackers pointing AI agents at more than 200 organizations.

The signal is plain enough. Watching agents and switching them off is turning into its own product category, and the biggest names in security would rather own it than watch someone else build it.