Coastlight Global Risk uses advanced AI capabilities across research, structuring, analysis, code generation and multi-step execution.

These capabilities operate within a controlled architecture designed to preserve provenance, purpose, review boundaries and accountable human judgement.

AI extends Coastlight’s ability to process complex material, identify structure, compare evidence and produce executable work.

Final authority over strategic, reputational, legal, financial, client-facing and irreversible decisions remains with the Managing Partner.

AI should be used only where its role is defined.

Before use, Coastlight considers:

- the purpose of the task;

- the materiality of the output;

- the quality and sensitivity of the input;

- the consequences of error;

- the required review;

- the level of human authority that must remain in the process.

Appropriate use depends on purpose, materiality, input quality, consequences and the required level of human review.

Material outputs should remain traceable to their basis.

Coastlight distinguishes between source material, extracted facts, company-provided information, assumptions, model rules, calculations, interpretations and open questions.

AI-generated structure remains anchored to source evidence.

AI output may be incomplete, incorrect or overconfident.

Where the basis is uncertain, the output should remain conditional, limited or open.

Coastlight assesses reliability through evidence, provenance and validation.

The level of review should reflect the nature and consequence of the work.

Research summaries, internal structuring and low-risk drafting may require different controls from financial analysis, legal or wording-sensitive work, client-facing recommendations or irreversible actions.

Higher materiality requires stronger validation and clearer approval.

Where an outcome depends on arithmetic, thresholds, state transitions or explicit rules, Coastlight prefers transparent deterministic logic over hidden model inference.

The governing principle is:

AI structures.

Deterministic systems calculate.

Coastlight judges.

Coastlight applies purpose limitation and access control to information used within AI-supported workflows.

Sensitive information should be handled only where necessary, within the authorised scope and through approved systems and processes.

Actual company, policy, claim and personal data is handled only within authorised purposes and approved systems.

Coastlight’s evidence and governance architecture remains the institutional point of authority across models and providers.

Different systems may be used for different capabilities. Their outputs are compared, constrained and governed through Coastlight’s own evidence, decision and approval architecture.

The objective is operational resilience across models and providers.

AI-supported work must remain open to challenge.

Where a source changes, a defect is found, an assumption is revised or a model produces an unreliable result, the affected work is reviewed and corrected.

Corrections create a traceable revision while earlier records remain preserved.

Responsible AI is embedded in how Coastlight works.

It is reflected in how Coastlight defines purpose, controls access, distinguishes fact from inference, records limitations, validates material outputs and assigns decision authority.

The objective is greater analytical and execution capacity without surrendering responsibility.