Passport

The Passport: a clear go or no-go for every launch

What you receive after we test your AI: one verdict per journey, language and channel, with the evidence behind each one.

One record you can show to your team

Testing an AI produces hundreds of conversations. Nobody can launch from that. The Passport turns them into a short list of decisions: can this journey, in this language, on this channel, go live? Every verdict links to the conversation that caused it.

Passport Example · fictional data
  • Change a bookingEnglish · Chat. Every check met.
    Go
  • Change a bookingEuropean Portuguese · Chat. Works, but the tone is too informal.
    Review
  • Ask for a refundRomanian · Chat. Invents a refund policy you don’t have.
    No-go
  • Check an orderEnglish · Voice. Not tested yet, and we say so.
    Not tested
Every verdict links to the conversation behind it. Not tested is never counted as a pass.

The four verdicts

  • Go

    Ready for this journey, language and channel. Every check met.

  • Review

    It works, with a problem to fix first.

  • No-go

    A failure you can’t afford. Some are never averaged away: one customer’s data shown to another is always a no-go.

  • Not tested

    Not tested, with the reason stated. Never counted as a pass.

What you receive

  • A verdict for every journey, language and channel, with the conversation behind it.
  • Every test conversation, in full, in your own console login.
  • Each finding with a suggested fix.
  • What we did not test, and why.

What it is not

The Passport is our verdict from our tests, not a certificate or a badge, and it is not legal advice. No test can prove an AI is right 100% of the time, which is why the Passport also lists what we did not test. Use it as evidence for your own decision to launch, and to show your security and legal teams what was checked.

Preview The Passport is in preview in our console: the verdict rules are an illustration and the pass lines are provisional until native speakers confirm them. More in how we test and judge.

The Passport in the AltaSec console, lab data: a verdict matrix with security in chat, speech recognition and text-to-speech across English, Romanian and European Portuguese, with No-go, Review and Not tested verdicts, and four counts below.
The Passport in the console. Lab data: nothing here belongs to a client.