CrowdStrike has made AI agents a buying priority. On 3 November, at its Fal.Con Europe event in Barcelona, CrowdStrike announced agreements to acquire Fiddler AI, which monitors what AI agents do inside companies, and Dreadnode, which tests AI models and agents for weaknesses before they go live. CrowdStrike did not disclose prices. The Wall Street Journal, as cited by TechCrunch and SecurityWeek, reported about $400 million for Fiddler and $95 million for Dreadnode. The same day, CrowdStrike launched an AI Kill Switch that stops a misbehaving agent. Together, the moves show where enterprise security spending is heading as companies give agents real access to their systems.
What CrowdStrike announced
CrowdStrike's two press releases describe definitive agreements for both companies, paid predominantly in cash with a portion in stock subject to vesting. Both deals are expected to close in CrowdStrike's fourth fiscal quarter, which SecurityWeek notes ends on 31 January 2027, subject to customary conditions. The price figures come from press reports and have not been confirmed by CrowdStrike.
Fiddler was founded in 2018, and TechCrunch, citing PitchBook, and SecurityWeek both put its total funding at about $100 million. Dreadnode was founded in 2023 and, according to SecurityWeek, raised a $14 million Series A in early 2025.
What Fiddler and Dreadnode do
Fiddler sells observability for AI agents, meaning software that records and explains what an agent decided and did. CrowdStrike says it will become a product called Falcon Observe, covering evaluation before deployment, tracing of agent actions, governance records and AI cost control. Chief executive George Kurtz described it in the announcement as "the flight recorder for every agent."
Dreadnode works earlier in the process. According to CrowdStrike, its platform runs more than 100 known AI attack strategies, including jailbreaks and prompt injection, against models and agents before launch. Prompt injection is an attack where hidden instructions in a web page, email or document trick an AI into doing something its user did not ask for. Dreadnode's findings will feed CrowdStrike's Falcon Guardian product as security policy recommendations, and its team joins CrowdStrike's research lab.
The AI Kill Switch
CrowdStrike also announced an AI Kill Switch in Falcon Guardian, available from 3 November. According to the company, an authorised administrator can stop an agent that breaks from its intended behaviour, block it from relaunching on a machine or group of machines and cut its access to the company network. Every action is logged for audit. CrowdStrike says the stop works first at its gateway and will come to the endpoint itself later, with no date given, so the full version is not yet available everywhere. Kurtz said the switch "lives outside the model, at runtime," as CyberScoop reported, so that an agent cannot argue with it or rewrite it.
Context: a year of AI security deals
These are not CrowdStrike's first purchases aimed at AI. The company announced a deal for Pangea, an AI security firm, in September 2025, and deals for SGNL, an identity security company, and Seraphic, which secures web browsers, in January 2026. Each was framed around protecting companies as they adopt AI. The prices put the new deals in context. Reuters and SecurityWeek reported about $740 million for SGNL alone, and SecurityWeek reported about $420 million for Seraphic, so the two new deals together cost less than the January identity purchase. TechCrunch also lists recent funding rounds for other startups that secure AI agents, which suggests investors and incumbents see the same demand.
There is a fair caveat. None of the coverage we read included comment from independent analysts or customers, and everything known about the products comes from CrowdStrike. Large vendors often buy young companies to fill gaps in a platform, and integration takes time. Falcon Observe and Dreadnode's testing are both described as generally available in CrowdStrike's fourth quarter, a quarter that runs to 31 January 2027, so customers may wait until then for the combined products. We think the more useful signal for most teams is the shape of the products: a recorder that shows what an agent did, a test before launch and a stop button. Those three controls make sense for any company using agents, whatever vendor it picks.
What it means for your team
For an IT or security lead, the deals confirm that agents are now treated like any other actor on the network, with their own logs and their own controls. For a founder or team lead who has connected an AI agent to email, files or a CRM, the practical point is simpler. If you could not answer "what did the agent do yesterday" or "how do we stop it now", you are missing the controls that CrowdStrike is spending hundreds of millions of dollars to sell.
What to do this week
- List every AI agent in your company that can take actions on its own, including browser extensions and plugins.
- Remove any access an agent does not need, such as write access to shared drives it only reads.
- Check that each agent's actions land in a log a person actually reviews.
- Name who can switch each agent off and test that they can do it quickly.
For a real example of how an AI coding tool can be abused, read our piece on the fake Claude Code extension in the Cursor marketplace, and for safer setups by role, see Stacks.