AI Use Policy

84EM uses AI tools as assistive development tooling — similar to an editor or framework — to improve speed and quality of engineering output.

Current AI providers

Additional AI-powered tools may be used in the development environment (editor integrations, MCP servers, image search). A current inventory is available on request.

We use commercial / API tiers where the option exists, with training opt-out configured where available. Provider-side data handling is governed by each provider’s data processing terms.

Human review

AI-assisted output is reviewed for material correctness before being committed or delivered. Review depth is calibrated to risk: security-sensitive code, database operations, and client-facing deliverables receive line-by-line review; routine refactors and well-tested patterns receive lighter review.

84EM is a solo practice; the principal engineer (31 years of web development experience) performs all review.

Client data handling

We do not include client confidential information, personal data, or NDA-restricted material in AI prompts without client approval. When client data is necessary and approved, we limit it to the minimum required for the task.

When client data is included in AI prompts, the AI provider acts as a subprocessor for that processing.

Autonomous AI workflows

Internal operational tasks (knowledge base enrichment, monitoring, summarization) use AI workflows that run without per-task human review. These workflows operate on internal 84EM data and on client metadata pre-approved for AI processing (project status, ticket summaries, time entries). They do not produce client-facing deliverables.

Client-facing code, content, and communications are produced under engineer review.

Data retention

Local development context (conversation history, session logs) is retained on 84EM systems per the development tooling defaults and is not centrally archived.

claude-mem retains compressed memory of development sessions to support continuity across conversations. It runs on internal 84EM infrastructure. Client confidential information is excluded from memory ingestion through session scoping; client metadata used for project tracking may be retained.

Provider-side telemetry and retention follow each provider’s terms. We select settings that favor data minimization where configurable.

Code provenance

AI-generated code is reviewed before commit. We do not knowingly accept AI output that reproduces identifiable third-party copyrighted code. Open-source dependencies introduced through AI-assisted work are reviewed for license compatibility on the same basis as manually selected dependencies.

Prohibited uses

AI tools are not used at 84EM to:

Subcontractors

When subcontractors or agency partners contribute to client projects, 84EM requires them to handle client data under confidentiality terms consistent with our own. Their specific AI tooling and policies are their own, and 84EM does not warrant them. Subcontractor involvement is disclosed to clients before work begins on engagements where subs will have access to client systems or data.

Provider incidents

If an AI provider discloses a security incident affecting our usage of their service, 84EM assesses potential client impact and notifies affected clients within 72 hours of confirmed impact.

Client controls

Clients may at any time: