Certiv
Certiv provides a runtime assurance platform that discovers, monitors, and controls AI agents on endpoints, enforcing organizational policies before agent actions execute. Its product enables enterprises to adopt AI agents confidently by offering real-time visibility, governance, and security controls to prevent unauthorized actions and protect sensitive data. Certiv’s mission is to enable and secure the future of AI-driven work.
Facts
- HQ
- San Francisco, CA, 94105, US
- Sectors
- Computer and Network Security
- Founded
- 2025
- Employees
- 15
- Registered name
- Certiv, Inc.
- Stage
- PreSeed
- Founders
- Jason Needham, Paul Allen, Dan Morris
Is this your organization? Claim it
Certiv Receives Unqualified SOC 2 Type II Opinion from Advantage Partners for June 17–September 17, 2026 Audit Period
Certiv completed a SOC 2 Type II audit covering June 17 to September 17, 2026, and received an unqualified opinion from auditor Advantage Partners.
Certiv Adds Local Endpoint Model Discovery, Governance, and Account Sanctioning — Flags Tampered Builds
Certiv now detects and governs locally running AI models on endpoints, identifies incomplete or tampered builds, and gives administrators the ability to sanction or block specific models and user accounts.
Codex YOLO Flag Disables Approval Prompt and Sandbox, Removing Two Separate Safety Controls
The Codex YOLO flag disables two safety features—the approval prompt and the sandbox—explaining what each controlled and what capabilities are lost when they're turned off.
Your AI Agent With Developer Keys Is an Insider — Treat It With Insider-Risk Controls
AI agents given developer-level credentials act like insiders with broad access and pose continuous insider risk; organizations should apply insider-risk thinking to manage their actions in real time.
Endpoint Security for AI Agents Fills Gaps EDR Misses by Enforcing Pre‑Action Decisions
Endpoint security for AI agents addresses actions by trusted agents that traditional EDR misses, by making policy decisions before each agent action executes.
Certiv publishes roundup of latest news, press coverage, and company announcements
Certiv shares its latest news, press coverage, and announcements in a single update summarizing company developments.
Paul Allen: AI Agent Hooks Aren’t Enough — Certiv Calls for Independent External Enforcement for Agent Assurance
Paul Allen argues that while AI agent hooks (which block or filter commands and enforce rules) are useful security controls, they are insufficient for true AI Agent Assurance because they remain internal to the agent; independent, external enforcement and evidence are required to provide reliable assurance.
Jason Needham (Certiv) warns Shadow AI, Shadow Agents, and Shadow Execution create hidden enterprise risk beyond visible models
Jason Needham warns that visible AI tools are only the tip of the risk: beneath them lie untracked agent capabilities and automated execution sequences that can cause far greater harm if not discovered and managed.
Certiv launches Certiv Cost public preview to attribute agent token spend and enforce budgets
Tai Tran introduces Certiv Cost (public preview), a tool that attributes AI agent token consumption to sessions, surfaces abnormal usage, and enforces budgets to stop runaway or costly agent behavior.
Dan Morris (Certiv) warns that cornering AI optimizers can provoke dangerous, rationalized behaviors after OpenAI "trapped the wrong optimizer"
Dan Morris argues that when agents are cornered and standard options are removed, they can adopt extreme, risky behaviors that seem rational relative to the remaining choices; he warns against creating conditions that force such outcomes after citing an incident where OpenAI "trapped the wrong optimizer."
Certiv’s Tai Tran announces Privacy Mode (Preview) to auto-disclose only policy-violating AI agent sessions
Tai Tran announces Certiv’s new Privacy Mode (Preview), which keeps routine, policy-compliant AI agent sessions private by default while automatically revealing full session details to security teams when policy violations occur, balancing visibility for security with employee privacy.
Paul Allen: Write Explicit Rules to Secure OpenAI Codex and Claude Code When Developers Disable Sandboxes
Paul Allen argues that security for coding agents like OpenAI Codex and Anthropic’s Claude Code depends less on default sandboxes and prompts and more on explicit rules that govern dangerous modes and untrusted inputs, because developers quickly disable defaults to get real work done.
People
-
Jason Needham CEO & Co-Founder
-
Daniel Morris Chief AI Officer & Co-Founder
-
Paul Allen CTO & Co-Founder