Disclaimer

Important limits on Beacon research output and AI-assisted reasoning.

Effective May 4, 2026. Beacon is a research agent, not an authority system. This page states the core output, governance, and reliance limits that users should understand before treating any generated report as decision support.

Search docs, support, and public pages
search
Output Limits

Research output can be incomplete, stale, or wrong

Evidence-driven, not certainty-driven

Beacon synthesizes from retrieved search results, source snippets, stored memory, and model reasoning. That means its outputs depend on source quality, retrieval coverage, and model behavior rather than ground-truth certainty.

Model error remains possible

Beacon can misread, over-compress, or overstate source material. It can miss nuance, fail to surface contradictory evidence, or produce a confident summary that does not fully match the cited material.

Memory can preserve old framing

Because Beacon stores prior run context and reuses it in later runs, an early misunderstanding or skewed source mix can influence future research unless a user actively checks and corrects it.

Sources

Source links help audit the output, but they do not eliminate risk

Source ledger is a review aid

Beacon is designed to show URLs, snippets, and report citations so a user can audit where claims came from. Those citations improve traceability, but they do not guarantee that the synthesis is complete or that the source itself is reliable.

Primary material should win

If the synthesized output conflicts with the linked source, the source should be treated as the stronger authority. Users should read original materials directly before relying on an important conclusion.

High Stakes

Do not use Beacon as a sole basis for consequential decisions

No legal, medical, or financial reliance

Beacon output should not be treated as legal advice, medical advice, financial advice, compliance advice, or any other regulated professional determination.

No automated rights-impacting decisions

Beacon is not intended to make or recommend final decisions about employment, credit, insurance, housing, education, benefits, law enforcement, safety controls, or a person's legal rights without meaningful human review.

Sensitive-data caution

Users should avoid placing sensitive personal, regulated, or confidential information into Beacon unless they have confirmed that the specific deployment and governance posture is appropriate for that risk level.

Review

Human review is mandatory

Independent verification required

Users must independently verify material facts, check source quality, and apply domain judgment before using Beacon output in any business, legal, operational, or public-facing context.

Decision responsibility stays with the user

Beacon may help frame a question, reduce search time, or summarize source material, but it does not assume responsibility for downstream choices. Accountability for decisions remains with the person or organization using the output.