Boston University professors Jessica Silbey and Woodrow Hartzog propose a narrow, usable definition of AI slop that focuses on effort, who bears the cost, and whether repeated use corrodes a field.
AI slop is defined by three measurable components: negligible exertion, asymmetrical imposition, and domain degradation — not by subjective quality alone.
The test treats those components as a spectrum, so outputs can be more or less sloppy rather than forcing a binary classification.
Regulatory timing matters: transparency duties under EU Article 50 and California’s AI Transparency Act began applying Aug. 2, 2026, sharpening enforcement questions.
Two Boston University law professors have given policymakers something the AI slop argument has lacked: a test with edges. Jessica Silbey and Woodrow Hartzog provisionally define slop as machine output produced with little exertion that shifts the burden onto recipients and erodes the domain it lands in. The test turns on effort, imposition and domain degradation rather than on quality, and that is what makes it usable.
Whether detection tooling hardens into an enforcement layer, carrying its false-positive problem. Providers of systems intended to interact directly with people must design and develop them so the person is informed they are interacting with an AI system, unless that is obvious to a reasonably well-informed observer. Under Section 22757.3(c), a covered provider that finds a licensee has modified its system so it no longer includes the required latent disclosure must revoke that license within 96 hours.
Meanwhile, the courts, where the counting has actually been done, are what the paper's policy catalog never reaches. Whether governance programs start treating unattributable AI output as a retention and defensibility question rather than an HR one. 6, 2026, timestamp in its running header and is labeled a draft on its interior pages, meaning the language quoted here may change before publication.
A consensus is emerging in law schools across Canada and the US that AI has no place in teaching the core curriculum, because it impedes student learning, especially in first year. Schools are beginning to follow the University of Chicago’s lead in banning AI in first-year core courses. Some profs are also banning lapt
From the Financial Times, although describing the practices of only two law schools. These days, we are all struggling to balance the human and the machine in the workplace. Take the training of lawyers. Generative AI can already do much of the grunt work normally given to young, inexperienced attorneys, but if we let
Josh Blackman (South Texas), AI Freezes The Scholarly Voice: I recently attended a workshop about how law professor are using AI. One professor said that he has trained AI on his writing style by reviewing all of his past publications. Thus, any text generated by the AI will be in the same sort of style […] The post Bl
Open the app view to save this story, compare related coverage, and continue from the same source.