Simaril
Silmaril is a self-healing prompt injection defense designed to protect AI systems from increasingly complex, multi-step attacks that traditional guardrails fail to stop. It operates by wrapping inference calls and evaluating whether an execution sequence is leading toward a harmful outcome, rather than simply filtering inputs. It uses a multihead classifier that analyzes user intent, application context, and execution states together, enabling it to detect indirect injection, multi-turn attack chains, context poisoning, and tool abuse before damage occurs. Silmaril continuously strengthens its defenses through autonomous threat hunting agents that probe systems, discover vulnerabilities, and generate synthetic training data from real attack scenarios. These insights are used to retrain the model automatically, deploying updated protections in under an hour and propagating anonymized defenses across all deployments.
Learn more
CanaryVaults
CanaryVaults is a deception-based security platform for AI applications and agents. It uses decoy signals to reveal exposed credentials, data leaks, prompt injection, and suspicious agent activity. Its six components include decoy email identities; CanaryRAG, which embeds invisible, plausible false facts in documents to reveal when their contents surface in AI responses; CanaryShield, a three-layer prompt-injection firewall available as an OpenAI-compatible proxy or an API for scoring text; CanaryHoneypot, decoy AI endpoints that return synthetic data and record attacker activity; CanaryAudit, a tamper-evident log of agent actions; and CanaryAgent (early access), which provides heartbeats, skill scans, and an action timeline for OpenClaw agents. Alerts are available through email, Slack, Discord, Telegram, and HMAC-signed webhooks. API access is included on all plans. A free plan requires no credit card, and paid plans start at $12/month.
Learn more
ZeroLeaks
ZeroLeaks is an AI prompt security platform that helps organizations identify and fix exposed system prompts, internal tools, and logic vulnerabilities that could allow prompt injection, prompt extraction, or other forms of leakage that expose internal instructions or intellectual property to unauthorized actors. It provides an interactive dashboard where users can scan system prompts manually or automate scanning via CI/CD integration to catch leaks and injection vectors before code is deployed, and it uses an AI-powered red-team-style analysis engine to assess prompt surfaces for logic flaws, extraction risks, and potential misuse with evidence, scoring, and remediation recommendations. ZeroLeaks targets enterprise-grade security for large-language-model-based products by offering vulnerability assessments that highlight prompt exposure depth, prioritized risks, proof, and access paths for issues found, and suggested fixes such as prompt restructuring, tool gating, etc.
Learn more
HOL Guard
HOL Guard is a local-first runtime security layer for AI agents that watches what an AI assistant is about to do and stops risky actions before they happen. It sits between the agent and the computer, evaluating supported tool calls and local artifacts for threats such as secret and credential exposure, destructive commands, prompt-injection-driven actions, malicious or changed packages, risky MCP configuration, and unsafe plugins, skills, hooks, and settings. Known threats can be blocked automatically, while ambiguous actions are paused for user approval so people remain in control. Guard runs entirely on the developer’s machine, works offline, and does not upload files, prompts, or passwords. Local checks typically complete in under 50 milliseconds and require no changes to existing code or routines. It supports coding agents including Claude Code, Cursor, Codex, Gemini CLI, OpenCode, Hermes, and OpenClaw, with tailored integrations that inspect actions before execution.
Learn more
