When AI systems built to act like humans cause harm, our legal system faces a question it wasn't designed to answer: who is responsible? The dominant response has been to treat AI as a passive instrument wielded by a human operator (what I call the Instrumental Frame), a logic embedded in product liability law long before Social AI existed. In Garcia v. Character Technologies, the Frame did essential work: the Court refused to let Character.AI dissolve the chatbot into an inert vessel, held the company accountable for its design choices, and compelled concrete safety improvements. But the Frame strains. Social AI isn't a passive object -- it acts back, by design -- and treating it as one carries real risks, including the risk that accountability folds into the nearest human in ways that protect the system itself at precisely the moment we need to be asking harder questions about what it is. My paper names this tension but does not resolve it. That resolution is a question for the judiciary, legislators, and society at large. For now, I suggest that the Instrumental Frame is a fiction worth keeping. What we do with the time it buys us is the harder question.

When ChatGPT started "breathing" during voice interactions this year, this design choice seemed innocuous. However, it highlights a concern around increasingly human-like AI features being implemented without proper evaluation of their necessity or impact. The tragic case of Character.AI -- where anthropomorphic design choices allegedly contributed to a teenager's death -- demonstrates the urgent need for systematic analysis of these features. This article proposes a two-step framework for evaluating anthropomorphic AI design: first assessing both social and technical necessity, then implementing appropriate guidelines and safeguards (some of which are suggested herein). While some applications (like accessibility tools) may well justify human-like features, others (like companionship AI) tread on ethically difficult territory. As AI becomes more sophisticated, we need clear criteria for determining when anthropomorphic design serves legitimate social purposes. An LLM doesn't really need to breathe, and perhaps we ought to think much harder about how we would like human-to-human relationships to look like in the future before delving too far into enabling human-to-AI facsimiles.

This week, a BBC Panorama investigation uncovered how troubling deepfake images of Trump surrounded by Black supporters found their way across the internet -- yet another example of why A.I. election disinformation regulation is a necessity in the US. In the courts, Microsoft filed a motion to dismiss part of the New York Times' case against it and OpenAI. The case alleges that ChatGPT imperils the very future of journalism. Microsoft compares The Times to Hollywood-types in the 1980s panicking about the doomsday device that was the VCR. Speaking of OpenAI, the company clapped back at Elon Musk's lawsuit by posting a bunch of Musk's emails on its blog which, frankly, leaves Elon looking kind of like a hypocrite. The company also signed an open letter pledging to build AI that will make for a better future for humanity. If those platitudes don't leave you feeling warm and fuzzy, don't worry. It isn't just you. Finally, India signalled a move towards active A.I. regulation with a much-criticised and extremely confusing advisory which requires companies (but apparently not startups!) to obtain government clearance before releasing new unreliable models (i.e., every single model) onto the Indian internet.