This is a neutral, unsponsored list of startups focused on AI security: securing models, agents, data paths, and identities. It's not a ranking or an endorsement. I follow these startups to keep up with new developments in AI security.
Join the AI Security community on Twitter and LinkedIn group for additional updates. Source: Awesome AI Security
Cranium
Enterprise AI governance and security: inventory AI systems, test for vulnerabilities, manage risk, and attest compliance across first- and third-party AI.
X · Blog · LinkedIn · Crunchbase
Fiddler
AI observability and guardrails: evaluate, explain, and monitor LLMs/agents in real time to reduce risk and improve reliability.
X · Blog · GitHub · LinkedIn · Crunchbase
HiddenLayer
Security for AI models and agents: supply-chain scanning, runtime defense, posture management, and automated red-teaming.
X · Blog · GitHub · LinkedIn · Crunchbase
Protecto
Privacy guardrails for GenAI: discover and mask PII/PHI while preserving context so AI apps stay useful and compliant.
X · Blog · GitHub · LinkedIn · Crunchbase
Giskard
LLM security and testing: continuous red-teaming to find jailbreaks, hallucinations, and safety flaws before and after deployment.
X · Blog · GitHub · LinkedIn · Crunchbase
Zenity
AISPM and agent security: govern and secure AI agents with posture management, policies, and threat prevention across environments.
X · Blog · LinkedIn · Crunchbase
Deepchecks
Evaluation and monitoring for LLM apps/agents: define metrics, run tests, and add safety checks to ship reliable AI.
X · Blog · GitHub · LinkedIn · Crunchbase
Astrix Security
Non-human (machine) identity and AI agent access security: discover, govern, and enforce least-privilege with full audit trails.
X · Blog · LinkedIn · Crunchbase
Descope
Agentic identity and auth for AI apps/agents: manage lifecycle, authentication/authorization, and policy guardrails for agent identities.
X · Blog · GitHub · LinkedIn · Crunchbase
Pillar Security
LLM/agent security across the lifecycle: fingerprinting and inventory, automated red-teaming, and enforceable guardrails.
X · Blog · LinkedIn · Crunchbase
Lasso Security
GenAI security platform: monitor AI use, detect risks in real time, and add runtime defense/guardrails for apps, employees, and developers.
X · Blog · GitHub · LinkedIn · Crunchbase
TrojAI
Secure-AI platform: red-team models at build time and protect apps/agents at run time to mitigate prompt attacks, data leaks, and unsafe outputs.
X · Blog · GitHub · LinkedIn · Crunchbase
Noma Security
Agentic AI security: discover agent risks, monitor behavior, and enforce safeguards to adopt AI agents safely at scale.
X · Blog · LinkedIn · Crunchbase
Jit
Agentic AppSec/ASPM: AI agents prioritize vulns, create risk assessments and threat models, and help drive remediation in developer workflows.
X · Blog · GitHub · LinkedIn · Crunchbase
SPLX AI
End-to-end AI security: scalable red-teaming, runtime threat detection/response, and automated governance and remediation.
X · Blog · GitHub · LinkedIn · Crunchbase
Harmonic Security
Data security for AI adoption: real-time visibility into AI usage, sensitive-data detection, and "human-like" protection and coaching.
X · Blog · LinkedIn · Crunchbase
Knostic
Enterprise AI security and compliance: need-to-know access controls for copilots/LLMs and policy enforcement to prevent oversharing.
X · Blog · LinkedIn · Crunchbase
Clutch Security
Universal non-human identity security: discover, govern, and enforce zero-trust for API keys, service accounts, and agent credentials.
Blog · GitHub · LinkedIn · Crunchbase
WitnessAI
Secure AI enablement: visibility of AI use, policy controls, and protection for sensitive data and topics across the enterprise.
Blog · LinkedIn · Crunchbase
Straiker
Agentic AI security: detect prompt injection, hallucinations, and rogue behaviors with red-teaming and guardrails for AI apps.
X · Blog · LinkedIn · Crunchbase
Aurascape
AI-native security layer: real-time visibility of AI use, intent-based controls, and guardrails for embedded and shadow AI.
X · Blog · LinkedIn · Crunchbase
Mobb
AI-powered auto-fixes for code vulnerabilities to shrink security backlogs; research on AI-generated app risks and remediation.
X · Blog · GitHub · LinkedIn · Crunchbase
Mindgard
Automated AI red-teaming and runtime controls: map the AI attack surface, validate risk, and deliver AI detection and response.
X · Blog · GitHub · LinkedIn · Crunchbase
promptfoo
Open-source framework/CLI for testing and red-teaming prompts/agents; CI/CD-friendly automated evals.
X · Blog · GitHub · LinkedIn · Crunchbase
Confident AI
Built by the creators of DeepEval, engineering teams use Confident AI to benchmark, safeguard, and improve LLM applications, with best-in-class metrics and tracing.
X · Blog · GitHub · LinkedIn · Crunchbase
Irregular
Frontier AI security lab that adversarially evaluates advanced models for misuse and builds defenses before release.
X · Publications · LinkedIn · Crunchbase
Citadel AI
Model risk testing and governance: jailbreak tests, monitoring, and reports aligned to standards to minimize safety, security, and compliance risk.
X · Blog · GitHub · LinkedIn · Crunchbase
Dream
AI cyber defense for governments and critical infrastructure with real-time insight, predictive defense, and resilience.
Blog · LinkedIn · Crunchbase
Prime Security
Design-stage security assistant that automates design reviews and risk mitigation for engineering teams.
