Alan Rozenshtein

@arozenshtein
1.4K Followers
132 Following
473 Posts
Assoc prof @UofMNLawSchool. Ed @ConstComm Sr & book rev ed
@lawfareblog. Co-host @RatlSecurity. Term member @cfr_org. Former @TheJusticeDept. Mr. @HNeprash.
Websitehttps://alanrozenshtein.com
With the Senate about to pass the TikTok ban/divestment bill, some thoughts in @lawfare are on the FIrst Amendment implications. Short version: The level of constitutional scrutiny doesn't matter. And the main issue is Chinese influence, not data privacy. https://www.lawfaremedia.org/article/five-observations-on-the-tiktok-bill-and-the-first-amendment
Five Observations on the TikTok Bill and the First Amendment

The law’s survival doesn’t hinge on the level of constitutional scrutiny. And the main issue isn’t really data privacy; it’s Chinese influence.

Default
🚨 New paper! Thanks to the Yale Journal on Regulation Bulletin
for publishing my essay, "Interpreting the Ambiguities of Section 230," in which I argue that refocusing on Section 230 as a standard statutory interpretation problem rather than a policy debate over liability is the best path forward for courts. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4800612
Correcting Presidential Immunity's Original Sin

In both civil and criminal cases, presidents should generally receive qualified, not absolute, immunity for official acts.

Default

"[M]odel weights are not generally used to express ideas among individuals but, rather, serve primarily as instructions to order machines to act."

Alan Rozenshtein pushes back on the argument that model weights for AI are protected by the First Amendment. https://www.lawfaremedia.org/article/there-is-no-general-first-amendment-right-to-distribute-machine-learning-model-weights

There Is No General First Amendment Right to Distribute Machine-Learning Model Weights

Unlike source code, which humans use to express ideas to each other, model weights function primarily as machine-readable instructions.

Default
I have a new piece up on @lawfare arguing that the First Amendment does not generally protect the distribution of machine-learning model weights. This is an important legal issue as the government considers AI export controls. https://www.lawfaremedia.org/article/there-is-no-general-first-amendment-right-to-distribute-machine-learning-model-weights
There Is No General First Amendment Right to Distribute Machine-Learning Model Weights

Unlike source code, which humans use to express ideas to each other, model weights function primarily as machine-readable instructions.

Default
Section 230 watchers: does anyone know of a good piece that analyzes the legal (not policy) arguments for and against FCC jurisdiction over Section 230?
Some thoughts for @lawfare on what we mean when we talk about ChatGPT having "First Amendment rights." https://www.lawfareblog.com/chatgpt-and-first-amendment-whose-rights-are-we-talking-about
ChatGPT and the First Amendment: Whose Rights Are We Talking About?

If ChatGPT is granted First Amendment rights, it won’t be because we are convinced that it has attained human-like personhood.

Lawfare
Tech law/policy friends: what are some good examples of when the tech industry (in particular the big platforms) successfully lobbied Congress (e.g., SOPA, encryption) and when they tried and failed (e.g., SESTA/FOSTA)? Ultimately, are they good lobbyists?
RT @1Br0wn
Good to see this detailed analysis by @arozenshtein. We tried making some of these points to the @EUCouncil during the #DigitalMarketsAct final negotiations, to justify keeping the @Europarl_EN’s #socialnetworking #interoperability mandate. But key players seemed wilfully deaf 😡
A paper from @arozenshtein which could serve as a useful primer on federated moderation issues:

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4213674#maincontent