# AI attribution decay and mitigation research > Live situation record from CLSTR: https://clstr.news/situations/ai-attribution-decay-and-mitigation-research > Updated: 2026-09-06T08:42:00.000Z. Sources: 14. Developments: 3. Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) have identified ‘attribution decay’ in generative AI models. This phenomenon occurs as training datasets grow, making it difficult to trace specific outputs back to individual pieces of training data. The study indicates that removing specific images or even an entire artist’s body of work may result in no measurable change to the model’s output, complicating efforts to assign credit or responsibility. To address this, researchers developed a ‘diffusion ensemble’ architecture. This system utilizes multiple smaller components trained on different data subsets, allowing researchers to test model outputs without specific data by switching off relevant components. The research has intensified debates regarding intellectual property and authorship. As the connection between original works and AI-generated outputs becomes diluted, legal and ethical questions arise concerning the protection of rights for creators whose works are used in training. In the creative industry, some entities are navigating these shifts through ethical guidelines; for example, the Rotterdam-based agency The Phoney Club uses licensed AI twins of real models but maintains a policy of not retouching natural imperfections to acknowledge the human reality behind the digital likeness. Recent discussions have further highlighted the tension between AI as a tool and AI as a replacement. Tim O’Reilly suggests AI can serve as a ‘new medium of expression’ if guided by clear human intention, whereas author Ted Chiang remains skeptical about whether AI can increase quality in tandem with quantity. ## Timeline ### 2026-09-06: Artificial intelligence transforms creative workflows and labor dynamics Artificial intelligence is transforming creative industries through new tool integrations like Magnific and ChatGPT, while raising critical questions regarding labor, data worker exploitation, and creator 책임. 5 sources. https://clstr.news/cluster/artificial-intelligence-sparks-debate-over-art-authorship-and-attribution ### 2026-08-20: MIT researchers identify attribution decay in generative AI models MIT researchers discovered ‘attribution decay’ in AI, where large datasets make it nearly impossible to trace generated images to specific training data, and developed a ‘diffusion ensemble’ to study it. 4 sources. https://clstr.news/cluster/mit-researchers-identify-attribution-decay-in-generative-ai-models ### 2026-08-18: AI research identifies attribution decay in generative models MIT researchers have identified ‘attribution decay’ in AI, making it difficult to trace outputs to training data, while creative studios adopt ethical rules for using AI-generated models. 5 sources. https://clstr.news/cluster/ai-research-identifies-attribution-decay-in-generative-models --- Cite as: AI attribution decay and mitigation research. CLSTR, https://clstr.news/situations/ai-attribution-decay-and-mitigation-research