-
Notifications
You must be signed in to change notification settings - Fork 0
FAQ and Roadmap
Q: Is ARGUS calling people fake?
No. ARGUS scores profiles algorithmically and publishes the score with full reasoning. The language throughout is deliberate: "flagged for review," "high risk," "scored X/100 based on these signals." Visitors see the evidence and draw their own conclusions. ARGUS never says "this person is fake."
Q: Can someone use ARGUS to target a real person maliciously?
This is why every submission goes through human review before publishing. Nothing is auto-published. If a submission looks like a malicious targeting attempt, it gets rejected. Abusive submitters are banned. The human-in-the-loop approval step is the single most important protection against weaponization.
Q: What's the false positive rate?
At the default publication threshold (trust score ≤ 40, confidence ≥ 75%), estimated false positive rate is approximately 4–6%. This is displayed on every published page. We're publishing probabilistic analysis, not verdicts. A 5% false positive rate means roughly 1 in 20 flagged profiles may be a real person. That's why the dispute mechanism exists, and why we display confidence and false positive rate prominently.
Q: What happens if a real person's account gets flagged?
They submit a dispute with identity evidence. If the evidence meets Tier 1 or Tier 2 standards, the page is removed or updated. The process is designed to be fair and fast (14 business days). This is the same standard major platforms use for account disputes.
Q: Does the browser extension spy on my browsing?
No. The extension only activates on the specific platforms listed in the manifest (LinkedIn, Reddit, X, Instagram). It extracts handles from profiles you're already viewing and queries the ARGUS API for a score. No browsing history is stored locally or sent to ARGUS. The extension does not track which profiles you view — it only queries scores for profiles that appear on your screen.
Q: Can platforms block or sue ARGUS?
Platforms can block the scraping infrastructure, which is why Agent-Reach maintains updated connectors. Legally, scraping publicly accessible data has significant precedent supporting its legality (hiQ Labs v. LinkedIn, Sandvig v. Barr). ARGUS only accesses public data — information visible to any logged-out visitor — and does not circumvent access controls.
Q: Is this competitive with the platforms' own safety efforts?
No — it's complementary. Platforms like LinkedIn and Reddit have internal trust and safety teams, but their detection is invisible to users. ARGUS makes the analysis visible and consumer-accessible, which platforms generally cannot do for legal and UX reasons. We're filling the consumer gap, not replacing platform-level moderation.
Q: How is this free? What's the catch?
There is no catch. ARGUS is open source (MIT license). The detection models are open source or free-tier APIs. The hosting is Cloudflare free tiers. There is no advertising, no data selling, and no freemium gate. It's maintained by practitioners who think this infrastructure should exist in the world.
Q: Can I self-host ARGUS?
Yes. Fork the repo, create your own Cloudflare resources, and deploy. Every configuration is in wrangler.toml. See Getting Started.
Q: How do I get my profile removed if it was wrongly flagged?
Submit a dispute at argus.googleadsagent.ai/dispute. The form is linked prominently on every profile page. See Dispute System for the full process and evidence tiers.
- D1 schema, Worker API, detection pipeline
- Admin approval dashboard
- Public profile landing pages + SEO
- Community submission form
- Browser extension (LinkedIn, Reddit, X, Instagram)
- Dispute/opt-out system
- Evidence vault (R2)
- Google Indexing API integration
- Audio engine (voice cloning detection via Resemblyzer)
- Video engine (frame-level deepfake, SyncNet lip sync)
- Post/comment analysis (batch comment section scoring)
- Engagement velocity anomaly detection (viral post analysis)
- Watchlist & monitoring (weekly re-score with alerts)
- Embed widget (
<script>tag for any website) - Reporter toolkit (deeper API access for journalists)
- Bulk CSV analysis endpoint
- Network cluster pages (coordinated network reports)
- TikTok platform connector
- Bluesky / Mastodon connectors
- Full graph database migration (Neo4j or Cloudflare's upcoming graph offering)
- Real-time coordination detection (streaming, not batch)
- Browser extension Firefox support
- Mobile app (iOS/Android) with scan-on-tap
- Verified journalist API tier
- Multi-language support (Spanish, Portuguese, French)
- Affiliate/referral network fraud detection (LinkedIn-specific)
Open an issue with the enhancement label to request features. Most-upvoted requests get prioritized.