Responsible AI & Governance

1. System Character

composabilityPro, powered by the ComposabilityScores™ framework, is a deterministic decision-support infrastructure operated by Composability Ltd and its brands, composabilityPro and ComposabilityScores™. Scoring is driven by fixed, versioned policy rules. Identical inputs produce identical outputs. No manual overrides, tuning, or black-box models influence the assessment once submitted evidence has been captured.

2. Where AI Is Used

We use large-language-model (LLM) providers for a narrow, structured task: deterministic extraction of claims from the evidence you upload (pitch decks, financial documents, traction data). The LLM does not generate scores, decisions, or recommendations. All scoring, banding, and decision logic is performed by code inside our platform under fixed rules.

Where model output is used, it is structured, logged, and subject to our evidence-check layer. The scoring engine never sees free-form model output directly.

3. No Advisory Relationship

Outputs from composabilityPro and ComposabilityScores™ are informational and decision-support only. They do not constitute investment advice, a regulated financial service, a recommendation to buy or sell any security, or any form of financial, legal, or professional advice. Composability Ltd is not authorised or regulated by the Financial Conduct Authority.

4. Evidence Integrity

Scores reflect the evidence you submit. Incomplete, inaccurate, or outdated evidence will produce assessments that reflect those limitations. You are responsible for the accuracy of submitted evidence. Assessments are versioned, timestamped, and auditable — every score can be traced back to the exact evidence state that produced it.

5. Fairness & Bias

The scoring framework is published in plain language inside the product. Category weights, threshold bands, and mandatory-risk rules are visible to every user of the same tier. We do not apply individualised tuning that would benefit or disadvantage specific users, founders, sectors, geographies, or stages. When bias risks are identified (for example, a category that systematically under-weights a legitimate pattern), updates are rolled into a new versioned release of the framework.

6. Human Oversight

Scoring is algorithmic; governance is human. A named product owner reviews:

• Every material change to the ComposabilityScores™ framework before release.
• User-reported issues routed through the Support Messages channel.
• Audit logs of framework-version transitions.

Users can escalate concerns about an assessment via Support Messages. We do not privately re-score individual users in response — if a concern is valid, it results in a framework update that applies to all users equally.

7. Data Used for AI Processing

Evidence submitted to the platform is sent to model-provider APIs strictly for the deterministic extraction step described in §2. We contractually disable use of your inputs for foundation-model training where supported by the provider. We log the model version, prompt fingerprint, and extraction output alongside every assessment to preserve auditability.

For the full list of sub-processors, see Privacy Policy §6.

8. Prohibited Uses

You agree not to use composabilityPro to:

• Manipulate scoring outcomes through fabricated or misrepresented evidence.
• Reverse-engineer the policy rules or extract the framework for resale.
• Present the platform's outputs as advice, or as outputs of a regulated service.
• Automate high-volume assessments without a signed commercial agreement.

9. Changes to this Policy

We will update this page when the ComposabilityScores™ framework moves to a new major version, when a new sub-processor is added, or when the AI-processing scope changes. The Effective date above reflects the current version. Historical versions are available on request.

10. Contact

Questions about this policy? Email support@composability.pro. More on the entity behind the platform at composability.pro.