Pretense is an AI security solution designed for enterprise engineering teams to safeguard proprietary code when utilizing AI tools. It functions as an AI firewall that mutates sensitive code identifiers before they reach any large language models (LLMs), ensuring that real identifiers never leave the organization's network. This approach allows developers to leverage AI capabilities without compromising intellectual property. Pretense deploys in 30 seconds and integrates seamlessly with existing AI tools, offering a local-first, open-source solution with no cloud dependency.
Key Features and Functionality:
- Code Mutation: Deterministically replaces every identifier—such as functions, variables, and classes—with synthetic tokens, allowing LLMs to process functional code without exposing actual identifiers.
- Secrets Detection and Blocking: Identifies and blocks API keys, tokens, passwords, and personally identifiable information (PII) before they leave the local environment, covering over 30 major secret formats.
- Enterprise-Grade Compliance: Provides a full audit trail of every mutation, ensuring readiness for SOC2, HIPAA, GDPR, and PCI-DSS compliance, with immutable logs for security teams and regulators.
- Broad Compatibility: Works with all major AI tools and LLMs, including Claude, Copilot, Cursor, OpenAI, Anthropic, Gemini, and Ollama, as well as various integrated development environments (IDEs).
- Rapid Deployment: Installs via a single command, with no dependencies, and operates offline, ensuring that no data leaves the system.
Primary Value and Problem Solved:
Pretense addresses the critical issue of proprietary code exposure when developers use AI coding tools. Traditional data loss prevention (DLP) solutions often fail to understand code, leaving organizations vulnerable to inadvertent leaks. By mutating code identifiers before they reach AI models, Pretense ensures that sensitive information remains within the organization's control, allowing teams to harness AI's benefits without risking intellectual property. This solution provides real-time insights and analytics, enabling organizations to monitor and control AI interactions with their codebase effectively.
