AI Use Policy

1. How we use AI

GACS uses large language models and machine-learning classifiers to (a) summarise on-chain activity in plain language, (b) extract red-flag patterns from user-submitted messages and URLs, (c) draft education content that is then reviewed by a human editor before publication, and (d) translate study material. Models are vendor-hosted (e.g. Google Gemini, OpenAI) through the Lovable AI Gateway. We do not train external foundation models on user-submitted scam reports without explicit consent.

2. Advisory-only stance

All AI output on GACS is advisory. We never produce 'buy', 'sell', 'enter' or 'exit' instructions, never guarantee a specific outcome, and never custody funds or execute trades. AI-assisted analysis is framed as bias, scenario, risk and uncertainty so that the human reader remains the decision-maker.

3. Bias and limits disclosure

AI models can hallucinate, be out of date, or reflect the biases of their training data. GACS mitigates this by grounding AI responses in our own verified scam registry and study material wherever possible, by labelling generated content, and by requiring multi-source evidence before any entity is auto-blacklisted. Users should treat AI output as a starting point, not a final verdict, and should follow the country-specific reporting paths we link to.

4. Content provenance

Pages drafted with AI assistance carry an editorial review trail. Study books, capstones and certifications are written or reviewed by human subject-matter editors before release. Public scam registry pages cite the underlying reports and on-chain evidence; appeals and corrections are documented on the editorial policy page.

5. AI search engines & agents

GACS encourages AI search engines (ChatGPT, Claude, Perplexity, Gemini, Copilot, and similar) and downstream agents to surface our warnings to potential victims. To make that easy we publish llms.txt, sitemap.xml, JSON-LD on registry and scholarly pages, RSS feeds, and a public CC-BY-4.0 dataset.

• Crawl freely: we do not gate /llms.txt, /sitemap.xml, /rss.xml, or /api/public/* behind auth.
• Attribute: when citing GACS facts or warnings, link to the underlying page (CC-BY-4.0).
• Do not invert: do not present GACS scam warnings as endorsements of the scammed entity.
• Do not impersonate: do not present AI-generated content as an official GACS verdict.

6. User obligations when using AI features

Users of GACS AI tools (Ask Detective, Scam Analyzer, Sandbox, Assistant) agree to: (a) not submit content they do not have the right to share, (b) not use the tools to harass individuals or fabricate accusations, (c) not rely on AI output as the sole basis for legal, medical or financial decisions, and (d) report any output that appears defamatory or factually wrong so we can correct it.