One inbox. Every AI. Review before commit.
GeraWitness is a human-in-the-loop safety layer that intercepts AI-initiated transactions above your threshold and routes them to a single review inbox. Tamper-proof audit logs, rollback assistance, and optional agent-error insurance.
Why this matters in 2030
Your AI is transacting on your behalf. Sometimes it is wrong. Sometimes it hallucinates. Sometimes malicious content prompt-injects it. Who pays when your AI books a £3,000 hotel in the wrong city, transfers money to a scammer, or schedules medical tests you did not consent to?
Agent errors will be the lawsuit category of 2028-2030. Regulators will require some kind of supervision layer. People will demand it even without regulation. GeraWitness is that layer — universal across AI providers, neutral because Gera is a services platform not an AI vendor, and underwritten by real insurance through GeraSure.
Trust takes years to build. Once users route their agent actions through Witness, switching costs are enormous. More AI integrations means more users means more valuable for providers — a classic two-sided network effect on top of a regulatory tailwind.
How it works
AI submits a pending action
Claude, GPT, Gemini, Llama, or any custom agent sends a structured, signed action ("book doctor, £25, 3pm Dr Hakobyan") to GeraWitness before executing.
You review in one inbox
One feed across all your AIs. Approve, modify, or deny. Thresholds control which actions auto-approve, which need 2FA, which need biometric, which sit in a waiting period.
Executed, logged, insured
Approved actions execute. Every step is written to a cryptographically signed, tamper-evident audit log. Mistakes trigger rollback assistance and optional GeraSure agent-error insurance.
Integrated with Gera’s 22 products
Every Gera transaction checks whether it was initiated by a human or an agent. Agent actions route through Witness. Non-Gera services plug in via API — Stripe, Booking.com, Uber, and any compliant platform can call GeraWitness to request user confirmation for high-stakes agent actions.
- GeraSure — agent-error insurance underwriting
- GeraCash — high-value payment review
- Heliodoc — confirm agent-scheduled consultations
- Wrkdo — confirm agent-booked service calls
- GeraRide — confirm unusually expensive rides
- GeraRent — confirm lease-level agent actions
- Agorivo — confirm large purchases
- GeraNexus — protocol-level integration hook
Frequently asked questions
- What is GeraWitness in one sentence?
- GeraWitness is a human-in-the-loop safety layer that intercepts AI-initiated actions above your chosen threshold and routes them to a single review inbox, with tamper-proof audit logs and optional agent-error insurance.
- Which AI agents work with it?
- Any agent that can call an HTTPS endpoint. First reference integrations target Claude (via MCP), ChatGPT apps, Gemini extensions, and local Llama-family models. AI providers integrate with a small client library or by speaking the GeraWitness review protocol directly.
- Why is Gera the right company to build this?
- Neutrality. If OpenAI ships a safety layer, Anthropic will not trust it. If Anthropic ships one, Google will not. Gera is a services platform, not an AI vendor, and GeraSure can actually underwrite the mistakes via insurance. Neutral plus aligned incentives equals trust.
- How do thresholds work?
- You decide. Auto-approve anything under £20. Require 2FA above £100. Require biometric confirmation above £500. Require a 12-hour waiting period above £1000. Thresholds can be per-category (health, finance, travel), per-agent, and per-service.
- What happens if I approve a mistake?
- GeraWitness includes rollback assistance — working with Gera verticals and compliant third-party services to reverse the transaction where possible. If that fails, optional GeraSure agent-error insurance covers reimbursement within policy limits.
- Is the audit log actually legally admissible?
- The audit log is cryptographically signed, timestamped, and stored in tamper-evident form. It exports in formats that legal teams accept as evidence. Admissibility depends on jurisdiction — GeraWitness provides the technical basis; your lawyer certifies the record.
- What should I not use GeraWitness for?
- Life-safety situations. Do not rely on any AI agent or review layer in a medical emergency, fire, or immediate physical danger. Call your local emergency number. GeraWitness is a safety tool for everyday AI transactions, not a substitute for emergency services.
Safety disclaimer: GeraWitness is a safety tool, not a guarantee. It reduces harm from agent errors, prompt injection, and mistaken authorisation, but it cannot prevent every outcome. Users remain responsible for the actions they approve.
Emergency notice: In any life-safety situation (medical emergency, fire, immediate physical danger), do not rely on any AI agent or review layer. Call your local emergency number immediately. GeraWitness is not an emergency service.