AI Security: CrowdStrike Adds Kill Switch For Unauthorised Agents

MarketForces Africa

Source published: Site published:

Original article · فارسی

AI Security: CrowdStrike Adds Kill Switch for Unauthorised Agents

CrowdStrike has announced an AI Kill Switch in Falcon Guardian. It can stop unauthorised AI agents at the gateway and prevent them from running again across an organisation.

CrowdStrike says the update brings together rules for individual agents, network checks, and endpoint protection. This helps connect the instructions and messages an agent receives to the actions it carries out.

Guardian’s AI Gateway is now available to all customers as part of the product. It checks prompts and responses, tracks usage, and applies security rules across managed and unmanaged devices.

Tools such as Claude Code, OpenAI Codex, GitHub Copilot, and Cursor can use agent-level hooks. Other new features aim to protect against prompt injection, harmful instructions, sensitive personal information, and harmful AI skills and MCP servers.

Network checks on endpoints are now available for Windows, while the Mac version is in early access. CrowdStrike says using both network and endpoint information helps fix security gaps that network checks alone cannot cover. This is especially important when agents use local AI models that do not connect to a network.

Other planned improvements include access controls that check user identity and wider protection for cloud workloads, SaaS, and cloud-based agent systems.

CrowdStrike plans to buy Fiddler. This deal would add tools for watching and understanding how AI agents behave. These tools would help organisations test agents before putting them into use and understand why they take certain actions.

The deal still depends on meeting the conditions needed to complete the purchase. The expected new features depend on bringing Fiddler’s technology into CrowdStrike’s products.

CrowdStrike said its planned purchase of Fiddler AI, a leading company in AI agent observability, would help it offer the industry's widest range of AI agent security and safety features through its Falcon platform.

“Every organisation is racing to get more out of AI, and every week we see agents chain steps together, hit an obstacle, and take actions they were never authorised to take,” said Michael Sentonas, president of CrowdStrike.

“One control point isn’t enough to stop that. It takes the agent, the network, and the endpoint, together. This is how we make security the enabler for AI adoption, not the obstacle.”

Agent-level hooks can see the prompt, but they cannot see the model’s response or catch every action an agent takes. Network-only security tools can watch network traffic, but they cannot stop an action because the agent carries it out on the endpoint, not on the network.