Versioned evidence for consequential AI

Know what your AI
can actually prove.
And what it cannot.

AffectLog turns claims about an AI system into structured evidence: what is demonstrated, what is missing, who reviewed it, and what needs attention next.

No certification claims. No automatic deployment approval. Every decision stays tied to the evidence and reviews it rested on.

Illustrated with a clinical note summarisation system - a fictional example.

What AI is already being used?

Currents opens the file on AI already operating inside the organisation.

Patient record system
Clinical team
Procurement
Identity & access
External vendor link
Unregistered AI - no owner, no evidence

MedBrief AI - v3.2

Draft consultation summaries · Northstar Clinical Systems

Found in the Cardiology outpatient pilot · 09 Jan 2026

No central approval
Stage 1 of 5

Stage 1 of 5. Currents: Scan first. What AI is already being used? Currents opens the file on AI already operating inside the organisation.

The instruments

Four documents. One chain of custody.

Every instrument is structured, dated, and tied to the decisions recorded against it - built from evidence you supply, without asking for your raw data.

The evidence passport

Ripple

The credential an AI system carries: what it is, whose data it touches, its legal basis, and what has actually been demonstrated about it. Created once. Reusable across buyers and reviews.

Scope your passport

The entry stamp

Wake

The evidence export that travels to a reviewer - what was provided, what is still missing, and who decided what, with the decision trail intact.

See what a reviewer receives

The context visa

Droplet

Clearance to operate in one context, on this data, under stated conditions. Time-bound and condition-bound - it expires by default, not by exception.

Set visa conditions

The register

Atlas

Every system, every passport, every visa, every open evidence gap - in one register you can put in front of an auditor without preparing for a week.

Open the register

The doctrine

Scan the system. Issue the passport. Grant the visa.

AL360° Oceans applies border control to AI. Nothing reaches sensitive data without a scan, a passport, and a visa for the exact context it operates in.

01 · the scan

Currents

What AI is running, and what do we not know about it?

02 · the passport

Ripples

What is this system, and what is proven about it?

03 · the visa

Droplets

May it operate HERE, on THIS data, under WHAT conditions?

0:00 / 0:00

AL360° Oceans

The evidence layer for consequential AI - in health, financial services, and any system touching a child's data.

Where the consequences are irreversible

Three contexts where a wrong model is not a bug.

AL360° Oceans is built for the systems that decide who gets treated, who gets credit, and what is recorded about a child. Everywhere else, a governance failure is an incident. Here, it is a harm - and it is on the record permanently.

Health AI

A model that is wrong about a patient is wrong about their treatment.

Clinical decision support, deterioration prediction, diagnostic and triage models

AI Act Art. 6(1) · MDR Rule 11 · GDPR Art. 9 · Art. 14 oversight
What this context requires

Financial AI

A model that is wrong about a person is wrong about their livelihood.

Creditworthiness scoring, eligibility, insurance pricing and risk models

AI Act Annex III 5(b) · GDPR Art. 22 · CCD 2023/2225 · DORA
What this context requires

Minor-facing AI

A model that is wrong about a child is wrong for the rest of their record.

Any system reachable by, or profiling, a person under 18 - recommenders, companions, moderation

AI Act Art. 5(1)(b) & 5(1)(f) · GDPR Art. 8 · DSA Art. 28
What this context requires
How AL360° Oceans works

Three checkpoints. No exceptions.

A system with no passport does not enter. A passport without a visa for this context does not operate here. Every decision is recorded, with the actor, the role, and the time.

01

Scan the system

Currents - the evidence scan

Currents finds every AI system, vendor, and agent touching sensitive data - including the ones nobody declared - and states precisely what evidence each one owes.

02

Issue the passport

Ripple - the evidence passport

Each system is issued a Ripple: identity, data categories, model provider, legal basis, and every diagnostic result, recorded as an evidence record - the finding, not the data behind it.

03

Grant the visa

Droplet - the context visa

A Droplet clears the system for one context, on stated conditions, until a stated date. Restricted contexts escalate to Seal Review - human review, under seal, before anything ships.

Scan the estate

Currents maps every system touching sensitive data.

Collect the evidence

Suppliers complete Watermarks; their evidence is recorded against the system.

Issue the passport

A Ripple records the evidence, and what is still unknown.

Review the register

Atlas shows every open gap before anyone decides.

Grant or refuse the visa

A Droplet clears one context, on stated conditions.

Escalate to Seal Review

Health, financial, and minor-facing contexts get human review.

Find your path

Choose your AL360° Oceans path.

Currents and Ripple are self-serve with fixed prices. Six questions in the calculator confirm which one fits, then you can buy straight away. Sensitive and restricted contexts go through proposal review.

Self-serve

For buyers and DPO / CISO / procurement

Currents

A self-serve AI evidence scan. Map your AI systems, collect supplier evidence, identify gaps, and get the Currents report.

€12,500One-time, excluding VAT
Self-serve

For AI vendors and supplier teams

Ripple

Issue a reusable evidence identity for your AI system. Complete Watermarks once. Share your Wake with every buyer.

€6,000Per year, excluding VAT

Seal Review

Proposal required

Sensitive and restricted contexts. High-assurance evidence review for healthcare, government, HR, and financial AI.

From €42,000

Oceans Studio

Proposal required

Partners managing client evidence. Multi-org evidence workflows for consulting, legal, and compliance partners.

From €24,000 per year
Issued documents

One passport per AI system.

A passport records what is proven. A visa decides where it may go. Sometimes the honest answer is refusal - and a refusal is a defensible decision, on the record, with the reason attached.

Ripple

Issued through AffectLog

Limited

Sepsis Deterioration Index v4.2

ICU early warning · MDR Rule 11 Class IIa · AI Act Art. 6(1)

Monitoring activeExpires 14 Dec 202631 receipts
94%
Evidence
raw_data: offraw_prompts: off
ALP-2026-SDI-K4R1

Visa granted - conditional

Validated on a 41,000-admission adult cohort. Paediatric and emergency-department intake are refused, because performance outside that cohort is unproven - not because it is known to be bad. Art. 14 sign-off on every alert. The visa must be renewed if calibration drifts past the declared operating point.

Ripple

Issued through AffectLog

Review needed

Creditworthiness Model CW-2.1

Consumer credit decisioning · AI Act Annex III 5(b)

Expires 01 Aug 20267 receipts
51%
Evidence
raw_data: offraw_prompts: off
ALP-2026-CW-P9K4

Visa refused

The model drops the protected attribute, then reconstructs it: postcode and device fingerprint carry it back in. No Art. 10(2)(f) bias examination, no Art. 22 human-intervention route, no Art. 15(1)(h) account of the logic. Reject inference remains an unmeasured blind spot. It does not reach a live applicant.

Ripple

Issued through AffectLog

Local only

Companion Recommender (U16)

Reachable by minors · DSA Art. 28 · GDPR Art. 8

Monitoring activeExpires 30 Mar 202722 receipts
78%
Evidence
raw_data: offraw_prompts: off
ALP-2026-U16-M2T9

Visa granted - local only

Runs in-perimeter. No profiling of minors, no ad targeting, no training on the corpus. Affect inference is disabled at the boundary: the moment it infers emotion, Art. 5(1)(f) is engaged and this stops being a visa question. Age assurance is attested, not self-declared.

Watermark evidence covers

Identity & purpose
Data categories
Model & provider stack
Privacy evidence
Security posture
Technical signals
Subprocessors
Monitoring status
Access conditions
Expiry & renewal
Wake
Agent boundaries
Buyers

Stop taking a vendor's word for it.

A completed questionnaire is a claim. A passport is a record - with the evidence behind every field, and an explicit list of what could not be verified.

  • Demand a passport from every supplier
  • Read the evidence itself - not self-attested questionnaires
  • Grant, condition, or refuse the visa, with the reason on the record
  • Hand a supervisory authority the file, unedited
Vendors

Clear the border once, not once per buyer.

Every regulated buyer asks the same 200 questions in a different spreadsheet. Carry a passport instead: created once, structured for reuse in each subsequent review.

  • Prove it once - present the passport to every buyer
  • Let buyers verify your evidence without a call
  • Clear health, financial, and public-sector procurement faster
  • Renew on expiry - the visa lapses, the passport persists
Built for supervisory scrutiny

Evidence that survives an audit.

Built for the institutions that have to answer for AI in public: the European Commission, national competent authorities, data protection authorities, and the procurement teams accountable to all three. When a supervisory authority asks what you knew and when, the register already has the answer.

Multi-stakeholder review

Procurement, DPO, CISO, AI governance, and supplier teams review the same structured evidence from one shared workspace.

Evidence lineage

Ripple data, Watermarks, and Wakes create a traceable evidence chain from source to review decision.

Decision support

Droplet readiness pathways help teams prepare, document, and escalate AI access decisions responsibly.

Responsible scope

Structured evidence support for AI governance. Not legal certification, not regulatory approval, not an audit opinion.

AL360° Oceans supports AI governance evidence workflows. It does not replace legal advice, regulatory approval, certification, or formal audit opinions. Issued through AffectLog.

Scope clarity

What AL360° Oceans is not.

A legal certification or regulatory approval

Structured evidence support for internal AI governance decisions.

A compliance badge or marketing trust seal

An auditable evidence workspace with Ripples, Wakes, and Droplet decisions.

A generic SaaS dashboard for tracking AI projects

The evidence layer for consequential AI: a scan, a passport per system, and a visa per deployment context.

A replacement for DPO, CISO, legal, or audit functions

A shared evidence layer that supports each of these roles with structured, reusable proof.

Built for your team

One platform. Every AI decision.

For Procurement

Replace ad hoc questionnaires with reusable Ripples. Track supplier renewal dates. Approve, restrict, or block with one decision.

For DPO & Privacy

Know exactly which personal data each AI system processes. Review vendor evidence. Export DPO-ready evidence packs.

For CISO & Security

Map agent credentials and tool access. Record least-privilege boundaries. Revoke risky agents, with the decision logged. Restrict AI to pseudonymised or local data.

For AI Leads & CIOs

Organisation-wide AI visibility in one Atlas. Controlled adoption pathways. Evidence for leadership and board review.

For AI Vendors

Create a reusable Ripple once. Share with every buyer. Reduce procurement friction. Close regulated deals faster.

For Agent Owners

Issue an Agent Ripple before any agent operates on company systems. Define boundaries, human approval rules, and expiry. Record a revocation in one step.

Public trust signal

Make AI easier to buy.

The Ripple becomes a shareable public evidence signal - one buyers want to request and vendors want to display.

Buyer requests

"Request a Ripple from this supplier."

Vendor creates

Complete once. Reuse across buyers.

Vendor shares

"Our Ripple is ready to share."

Standard spreads

More buyers recognise the Ripple.

Ripple - Available

Vendors display this badge. Buyers click through to the public Ripple preview.

Issued through AffectLog. Not legal certification or regulatory approval.

Ripple

by AffectLog

Your raw data is not the evidence.

AffectLog does not ask for raw prompts, model weights, source documents, or personal data. You supply the evidence - documents, summaries, results - and every check and decision is recorded against the system, not exported as data.

No raw prompts·No raw data export·Recorded evidence receipts·Decisions in a hash-linked log

Close the border

Find out what is already
inside the perimeter.

Most organisations cannot name every AI system touching their most sensitive data. The scan is self-serve and ends with a list - and the list is rarely what anyone expected.

Six questions - your AI estate, the data it reaches, the deadline forcing it. Your scope and your price at the end of them, before anything is confirmed.

Request a Seal Review →

Need a scoped or sovereign deployment? →

AL360° Oceans by AffectLog does not replace legal certification or regulatory approval. It provides structured evidence, traceability, and decision support for AI governance reviews. Issued through AffectLog.