AI-powered · MCP-ready · UK-built

Pre-deployment security evidence for frontier AI.

Red-team results, threat models and hardening evidence for frontier and ASI-class systems — packaged as the security evidence pack regulators and customers demand.

Built for speed, compliance and scale

Frontier red-teaming

Adversarial, jailbreak and exfiltration testing scoped to ASI-class capabilities and weights security.

Threat model builder

Map attack surfaces, insider risk and model-weight theft into a structured, defensible threat model.

Hardening evidence

Capture controls, mitigations and residual risk as signed, verifiable evidence packs.

Customer trust exports

Ship security evidence packs to enterprise buyers and procurement teams on demand.

Verified live · 2 August 2026

Live trust spine

Only estate surfaces verified responding today are listed here. Each link was probed and returned a live response on 2 August 2026.

Free DEFONEOS public tool · works today

Sign and verify anything with Ed25519 — free, no account

The same sovereign signing backbone that seals ASISecurity evidence packs is open to everyone. Sign a message, get an Ed25519 signature and public key, then verify it offline. Both calls below were executed against the live API on 2 August 2026 and returned {\"ok\":true}.

1 · Sign

curl -X POST https://os.meok.ai/api/sign \\
  -H "Content-Type: application/json" \\
  -d '{"message":"my-evidence-pack-v1"}'

Returns signature, publicKey and a SOV: fingerprint.

2 · Verify

curl -X POST https://os.meok.ai/api/verify \\
  -H "Content-Type: application/json" \\
  -d '{"message":"my-evidence-pack-v1",
  "signature":"<hex from step 1>",
  "publicKey":"<hex from step 1>"}'

Returns {\"ok\":true,\"valid\":true} — reproducible offline, no account needed.

The safety surface family

Four sibling public surfaces, one evidence stack. No trackers, no dark patterns — just measured results and signed attestations, cross-linked so you can verify each layer yourself.

Connect via MCP

ASISecurity exposes its flagship tools as Model Context Protocol servers — call them directly from any MCP client, agent or workflow.

ai-incident-reporting-mcp.report

File and track AI security incidents directly from any agent or pipeline.

asisecurity-mcp.redteam

Launch automated red-team probes and collect signed evidence via MCP.

Try ai-incident-reporting-mcp.report

No installation required. This simulates the MCP tool call and shows the kind of output you can surface inside Claude, Cursor, Kimi or any A2A client.

  • Open-source MCP server
  • Local-first — no data leaves your machine
  • Signed audit trail on Pro+

Ready to make ASISecurity work for you?

Start a free trial, request a demo, or apply to become a partner.