Trust & Safety

The single most important non-feature in the product — a marketplace with weak trust mechanics doesn't retain users no matter how good the discovery/UX is.

Identity verification

• Tier 1 — Email/phone verified: baseline for any account, required before posting a gig or applying to one.
• Tier 2 — Identity verified: government ID upload + basic KYC check (especially important on the hirer side to prevent fraudulent payment activity, and on the creator side to enable payout without fraud). India: Aadhaar/PAN-based verification is the practical path.
• Tier 3 — Business verified (hirer-specific): registered business entity documentation, verified company domain email, or GSTIN — unlocks a "Verified Business" badge creators can trust before accepting an engagement.

Audience/follower authenticity

This is the hardest and most important creator-side trust problem — fake followers and bought engagement are endemic in the influencer space.

• v1 approach (pragmatic, no deep platform API dependency):
  • OAuth-connect social accounts where the platform allows it (confirms account ownership, pulls basic public stats) rather than pure self-reported numbers.
  • Engagement-rate sanity checks: flag profiles whose engagement rate is statistically implausible for their follower count (e.g., <0.5% or >15% sustained) for manual review, not auto-rejection — false positives are common, this should route to human review.
  • Community/report-based signal: hirers can flag a creator's audience as suspicious post-engagement, feeding a review queue.
• Phase 2: integrate a third-party audience-authenticity API (e.g., HypeAuditor-style fake-follower analysis) once volume justifies the cost — flagged as a build-vs-buy decision for later, not a v1 blocker.
• Never fully solvable — set expectations accordingly in policy copy. The goal is "materially better than DM-based sourcing," not perfect fraud elimination.

Payment & escrow protection

See 08-payments-escrow.md for full mechanics. Trust & Safety's role: fraud detection on payment methods (stolen cards, chargebacks), monitoring for money-laundering-pattern abuse (marketplaces are a known laundering vector — structuring payments through fake gigs), and enforcing KYC thresholds before high-value payouts.

Content & conduct policy

• Prohibited categories: adult content, gambling, counterfeit goods, medical/pharma claims requiring regulatory approval, political campaigns (jurisdiction-dependent), hate speech/harassment — a defined prohibited-category list gates what gigs/profiles can even be posted, enforced partly by automated keyword/category flagging and partly by report-driven human review.
• IP/content ownership disputes: policy for who owns what when — creator retains underlying content ownership by default; hirer receives a usage license as specified in the brief (see 03-features-brand.md usage rights tracking). Full copyright transfer must be an explicit, separately negotiated/priced term, never a default.
• Harassment/abuse reporting: any user can report another; repeated or severe violations escalate to suspension per 04-features-admin.md moderation tooling.

Dispute resolution policy

• Structured mediation, not arbitrary judgment calls: both parties submit their position with evidence (brief, delivered content, message log — all auto-attached from the platform's own records, which is exactly why keeping the brief/deliverables/communication in-platform matters so much).
• Default resolution principles (starting policy, to be refined with real case data):
  • Non-delivery past deadline with no communication → refund to hirer.
  • Delivery matches brief, hirer simply changed their mind → funds release to creator.
  • Delivery partially matches brief → partial split, admin discretion with documented reasoning.
  • Pattern of disputes against a single creator or hirer → account-level review, independent of any single case's outcome.
• Appeals path: one level of appeal to a senior reviewer before a decision is final.

Platform integrity monitoring

• Off-platform circumvention detection: messaging content-flagging for phone numbers/external payment requests in early-stage conversations (a common way marketplaces lose commission — parties agree on-platform then transact off-platform to dodge fees). Should be a soft nudge/warning in v1 (education: "keep it on-platform for payment protection"), not aggressive auto-moderation that creates false positives and user frustration.
• Fake account / bot detection: standard signals (velocity of account creation, device/IP clustering, disposable email domains).
• Review manipulation detection: flag review patterns suggesting fake/reciprocal reviews (e.g., a burst of 5-star reviews from accounts with no other platform activity).