# Cisco and Anthropic Strengthen AI Security for Enterprises

> The collaboration between Cisco and Anthropic introduces a groundbreaking approach to AI security, focusing on prompt intent to thwart potential threats. As organizations rapidly adopt AI technologies, safeguarding these systems becomes a top priority.

**Source**: blogs.cisco.com | **Published**: 2026-08-11 | **Type**: article

## Key Facts

- Cisco AI Defense enhances security by inspecting prompts, reducing AI manipulation risks significantly.
- Traditional data loss prevention fails to catch AI threats, exposing vulnerabilities in enterprise workflows.
- Anthropic's integration with Cisco provides a competitive edge in AI security, ensuring user trust and safety.
- Real-time protection without workflow changes boosts user adoption, enhancing financial performance for Cisco.
- The shift to AI-first security models indicates a broader market trend towards proactive risk management.

## Summary

Cisco has announced a strategic integration of its AI Defense capabilities with Anthropic's Claude Enterprise, aimed at enhancing security for organizations leveraging AI tools. This collaboration is significant as it addresses a growing concern in the AI landscape: the vulnerabilities associated with prompt injections and jailbreaks that can manipulate AI models and their outputs. As enterprises increasingly adopt AI technologies for various applications, ensuring the integrity and security of these systems has become paramount.

Claude Enterprise, which includes functionalities like Claude Code and Claude Cowork, allows users to interact with AI in ways that automate tasks, read documents, and interface with various tools. However, this capability also creates new attack surfaces. Traditional data loss prevention (DLP) systems, designed to monitor sensitive data leaving an organization, are ill-equipped to detect attempts to manipulate AI models. Cisco's AI Defense shifts this paradigm by focusing not just on data exfiltration but on understanding and evaluating the intent behind user prompts.

The integration utilizes Anthropic’s inference hooks, which serve as a checkpoint for every prompt submitted to Claude Enterprise. When a user inputs a prompt, it is sent to Cisco AI Defense for inspection before any action is taken. This system evaluates the prompt for potential threats, including prompt injections and jailbreak attempts. Safe prompts are allowed to proceed, while malicious ones are blocked, preventing any harmful actions from being executed. This proactive approach to security is a departure from traditional DLP methods, which primarily react to data leaving the system rather than assessing the conversation context itself.

Cisco AI Defense enhances security by monitoring not only the prompts but also the entire conversation transcript, including interactions with tools and the results of those interactions. This comprehensive oversight allows organizations to enforce security policies dynamically, ensuring that AI tools are used safely and effectively. As a result, security leaders can govern the use of Claude Enterprise without disrupting the user experience, maintaining the efficiency that users expect from AI technologies.

The implications of this integration are profound for the market. As AI adoption accelerates across industries, the need for robust security measures will become increasingly critical. Companies that can provide comprehensive security solutions will likely gain a competitive advantage. Cisco’s move signals a broader trend toward integrating security directly into AI workflows, rather than treating it as an afterthought. This shift could redefine how organizations approach AI governance, making security an inherent part of the AI lifecycle.

Looking ahead, the ongoing development of Cisco AI Defense, particularly as it expands to include response-side enforcement, will be crucial. As Anthropic continues to refine its inference hooks, the potential for real-time, inline security measures will grow, further embedding AI security into enterprise operations. This evolution suggests that organizations will not only need to adopt AI technologies but also invest in integrated security solutions that can adapt to emerging threats in a rapidly changing digital landscape. The future of AI in business will increasingly hinge on the ability to balance innovation with security, making this integration a critical step for enterprises navigating the complexities of AI deployment.

## Entities

- **Companies**: Cisco, Anthropic
- **Products**: Claude Enterprise, Claude, Claude Code, Claude Cowork, Cisco AI Defense
- **Technologies**: Model Context Protocol (MCP), inference hooks
- **Organizations**: Cisco Cloud Control

## Key Concepts

AI security, prompt injection, jailbreaks, data loss prevention, real-time protection, runtime policy, AI lifecycle, governed prompts

## Definitions

- **inference hooks**: Inference hooks allow organizations to send prompts to an AI security service for inspection before they are processed by the AI model.
- **prompt injection**: Prompt injection is a technique used to manipulate an AI model by embedding malicious instructions within a prompt.
- **jailbreak**: A jailbreak is an attempt to bypass the security measures of an AI assistant, allowing unauthorized actions or access.
- **runtime policy**: Runtime policy refers to the set of rules and guidelines that govern how AI models should process prompts and handle sensitive data.
- **data loss prevention**: Data loss prevention is a strategy designed to prevent sensitive data from being accessed, used, or transmitted inappropriately.

## Use Cases

- Inspecting prompts for AI threats before processing
- Blocking malicious prompts in real-time
- Governed AI interactions in enterprise settings
- Evaluating agent activity in conversation transcripts
- Integrating security measures without altering user workflows
- Providing a direct enforcement point for AI policy

## Frequently Asked Questions

**What is Cisco AI Defense?**

Cisco AI Defense is a security solution that inspects prompts sent to AI models for potential threats, ensuring safe interactions with AI systems. It integrates with tools like Claude Enterprise to provide real-time protection.

**How does the integration with Anthropic's inference hooks work?**

The integration allows Cisco AI Defense to inspect each governed prompt before it is processed by the AI model. This ensures that any malicious prompts are blocked, enhancing the security of AI interactions.

**What types of threats does Cisco AI Defense protect against?**

Cisco AI Defense protects against prompt injections and jailbreaks that could manipulate the AI model or its tools. It evaluates the entire conversation context to identify and mitigate these threats.

**Do users need to change their workflows to use Cisco AI Defense?**

No, Cisco AI Defense is designed to integrate seamlessly into existing workflows. Users can continue using Claude Enterprise without any changes while benefiting from enhanced security measures.

**What is the significance of runtime policy in AI security?**

Runtime policy is crucial as it defines how AI models should handle prompts and sensitive data during processing. It ensures that security measures are consistently applied to protect against potential threats.

## Links

- [Read on Welcome.AI](https://welcome.ai/content/cisco-and-anthropic-strengthen-ai-security-for-enterprises)
- [Original source](https://blogs.cisco.com/ai/secure-claude-enterprise-with-cisco-ai-defense)
- [Cisco](https://welcome.ai/company/cisco): Featured company

---

Source: Welcome.AI | https://welcome.ai/content/cisco-and-anthropic-strengthen-ai-security-for-enterprises