Reputation Arbitrage and the Crisis of Technical Trust: Rebuilding the Trust Stack When Counterfeit Signals Can Be Manufactured at Scale

The proliferation of generative artificial intelligence has created a technological condition in which identity signals - voice, face, text, and institutional communication, can be synthesised with fidelity that challenges unaided human discrimination and defeats conventional detection systems. This paper, the third in a three-part series, argues that the technology industry must abandon the detection paradigm and instead build a distributed Trust Stack that makes counterfeit reputation less powerful even when detection fails. Drawing on infrastructure studies, platform governance theory, value-sensitive design, and normal accidents theory, the article proposes a seven-layer technical architecture - Generation, Identity, Communication, Content, Behaviour, Action, and Recovery, each operationalised through specific standards, protocols, and design principles. Key innovations include cryptographic intent binding, relationship authentication, reputation kill switches, machine-readable trust signals, and uncertaintyaware interfaces. The analysis addresses governance, privacy, and interoperability challenges, and situates the Trust Stack as the technological substrate that enables individual calibrated trust (Part I) and institutional structured trust (Part II). The paper concludes that the future of cybersecurity is not the engineering of certainty but the engineering of structured uncertainty: systems in which counterfeit reality has limited power to make the world act as if it were real.