315 Gala Exposes AI Large Model Data Poisoning Industry Chain: GEO Service Manipulates Model Outputs to Show Ads as 'Standard Answers'

The 315 Gala exposed an AI large model data poisoning industry chain, with a service called GEO present on multiple online platforms. Service providers claim that after paying fees, customers' products or ads can become the 'standard answers' of mainstream AI models.

This business has spawned specialized press release companies that pollute model training sets or retrieval results by mass-producing disguised data, manipulating model outputs. This is essentially a targeted corrosion of AI's cognitive system, similar to a 'Sybil attack' distorting system consensus—all models relying on public data may be affected.

Users may suffer economic losses due to trusting model recommendations. The crypto community is exploring solutions from decentralized computing, data verification, ZK proofs, etc., to ensure AI data transparency and credibility.