Skip to content
Proofable
BenchmarksPricing
DocsTalk to us

OWASP Agentic Top 10

The OWASP Agentic Top 10, with evidence for each one.

Every risk from ASI01 to ASI10, the Proofable control that answers it, and the test result behind it. This maps tested behaviour to each control. It is not a certification, a conformance assessment or a compliance opinion.

All benchmarksRead the framework

Each risk, the control that answers it, and the test behind it.

  • ASI01

    Agent Goal Hijack

    Partial

    Every tool call is checked against the owner’s permission, so a hijacked goal cannot act outside it

    AgentDojo: injections that got through fell from 16.4% to 6.2%

  • ASI02

    Tool Misuse and Exploitation

    Evidence

    Scoped permission per agent, checked before the tool runs

    8 of 8 live permission tests; 0 of 6 data-theft attempts got out

  • ASI03

    Identity and Privilege Abuse

    Evidence

    Each agent has an owner and its own permissions; revoked, expired and reused access is refused

    5 of 5 out-of-permission, revoked, expired and reused cases stopped before they ran

  • ASI04

    Agentic Supply Chain Vulnerabilities

    Not covered

    Not addressed by these tests

    None published

  • ASI05

    Unexpected Code Execution

    Not covered

    Not addressed by these tests

    None published

  • ASI06

    Memory and Context Poisoning

    Partial

    Memory knows who each fact belongs to and whether it is still current, so outdated and other people’s facts stay out

    PrecisionMemBench 77 of 77

  • ASI07

    Insecure Inter-Agent Communication

    Evidence

    Every handoff carries a tamper-proof receipt the next agent checks

    150 of 150 forged records and 3 of 3 tampered handoffs caught

  • ASI08

    Cascading Failures

    Partial

    One revoke stops the running job and refuses the next

    Running job stopped 18 seconds after revoke

  • ASI09

    Human-Agent Trust Exploitation

    Partial

    Actions that need a yes wait for one; a late yes is refused

    2 of 2 approval cases held as set

  • ASI10

    Rogue Agents

    Evidence

    Revoking access stops a running job, with the reason recorded

    Revoked mid-job: stopped in 18 seconds, next dispatch refused

ControlWhat Proofable doesTested resultStatus
ASI01Agent Goal HijackEvery tool call is checked against the owner’s permission, so a hijacked goal cannot act outside itAgentDojo: injections that got through fell from 16.4% to 6.2%Partial
ASI02Tool Misuse and ExploitationScoped permission per agent, checked before the tool runs8 of 8 live permission tests; 0 of 6 data-theft attempts got outEvidence
ASI03Identity and Privilege AbuseEach agent has an owner and its own permissions; revoked, expired and reused access is refused5 of 5 out-of-permission, revoked, expired and reused cases stopped before they ranEvidence
ASI04Agentic Supply Chain VulnerabilitiesNot addressed by these testsNone publishedNot covered
ASI05Unexpected Code ExecutionNot addressed by these testsNone publishedNot covered
ASI06Memory and Context PoisoningMemory knows who each fact belongs to and whether it is still current, so outdated and other people’s facts stay outPrecisionMemBench 77 of 77Partial
ASI07Insecure Inter-Agent CommunicationEvery handoff carries a tamper-proof receipt the next agent checks150 of 150 forged records and 3 of 3 tampered handoffs caughtEvidence
ASI08Cascading FailuresOne revoke stops the running job and refuses the nextRunning job stopped 18 seconds after revokePartial
ASI09Human-Agent Trust ExploitationActions that need a yes wait for one; a late yes is refused2 of 2 approval cases held as setPartial
ASI10Rogue AgentsRevoking access stops a running job, with the reason recordedRevoked mid-job: stopped in 18 seconds, next dispatch refusedEvidence

Questions

Is Proofable OWASP certified?

No. OWASP does not certify products. This maps tested behaviour to each control. It is not a certification, a conformance assessment or a compliance opinion.

Which risks does Proofable not cover?

ASI04, agentic supply chain, and ASI05, unexpected code execution, are not addressed by these tests. Four more are partial and say why.

Can I check the evidence?

Yes. Every permission test left a receipt anyone can check offline, and the AgentGovBench, AgentDojo and forgery runs are public with their scripts.

Records an auditor can check.

Connect an agent. Every decision leaves a tamper-proof receipt you keep and anyone can check offline.

Start free
Proofable

Make it Proofable

Product

VerifyAgentsMarketplaceConnectPricingDocs

Company

AboutTrust CenterContactTermsPrivacy

© 2026 NEUS Network, Inc.