A2IS2E
How It Works

Simple to start, rigorous in delivery

Whether you are commissioning a red team, training your own experts, or reporting harm you have witnessed, the path is clear.

For Organisations

  1. Step 1

    Request

    Share your AI system details, deployment context, risk areas, and timeline. No technical red teaming knowledge required.

  2. Step 2

    Match

    We assemble a team with the right mix of languages, domain expertise, and red teaming experience. You approve the team before testing begins.

  3. Step 3

    Test

    Red teamers follow structured protocols adapted to your system. We coordinate logistics, secure access, and monitor progress.

  4. Step 4

    Report

    You receive a prioritized report with evidence, severity ratings, and recommendations. We can also provide a live debrief and optional retesting.

For Training Clients

  1. Step 1

    Needs Assessment

    Tell us about your team, your systems, and your sector. We recommend the right training format and duration.

  2. Step 2

    Design

    We tailor the curriculum to your context. Examples and exercises come from your sector.

  3. Step 3

    Deliver

    We run live online or in-person workshops, bootcamps, or the two-week twinning visit.

  4. Step 4

    Support

    After training, your team can access mentoring, templates, and refresher sessions.

For Whistleblowers and Reporters

  1. Step 1

    Submit a Report

    Use our Report AI Harm form to tell us about the system, the company involved, the use case, and the country where the harm occurred. You may remain anonymous.

  2. Step 2

    Initial Review

    Our team reviews the report within 10 working days. We assess the nature of the harm, public interest, and available evidence.

  3. Step 3

    Investigation or Pro Bono Evaluation

    If the case meets our criteria, we may offer pro bono red teaming or connect you with partners who can investigate further. We will never share your identity without your consent.

  4. Step 4

    Public Interest Action

    Where appropriate and with your agreement, we may publish a redacted public interest summary to raise awareness and encourage remediation.

Open the Report AI Harm form

Security and Confidentiality

  • All red teamers sign NDAs and agree to responsible disclosure.
  • Findings are shared only with the client through secure channels.
  • Whistleblower identities are protected under our confidentiality policy.
  • We can work within your existing security review processes.
  • Data protection and privacy are built into every engagement.

Start with a conversation

Tell us what you are building or what concerns you. We respond within 48 hours.

Contact A2IS2E