This is a regulatory briefing memo from a law firm tracking FTC enforcement trends. The source describes "algorithmic disgorgement"—court-ordered deletion of AI models trained on illegally collected data—as an emerging FTC remedy. The authors document the doctrine's evolution from theory (2019 Cambridge Analytica settlement) through multiple applications, culminating in the December 2023 Rite Aid case, which marks the first time the FTC used Section 5 unfairness authority to target discriminatory AI deployment. The Rite Aid settlement requires deletion of facial recognition models and underlying datasets due to racial misidentification harms. For practitioners, the operative takeaway is concrete: the FTC now treats illegally-sourced training data as poisoning the entire model artifact, not just requiring disclosure or modification. An FTC official quoted in the source calls this a "significant part" of AI enforcement strategy going forward. The practical implication is that companies building or acquiring models face real destruction risk if data provenance cannot be documented as lawful, particularly where the model produces discriminatory outputs. The source doesn't detail how courts will assess whether deletion is proportionate, or whether this applies retroactively to models built before the doctrine solidified.