🧵 I want to crack down on Big Tech as much as anyone, prob more than most. But the draconian and unworkable kids' online safety bill that the MA House passed earlier this month is not the way to do it. From @evangreer.bsky.social & me in today's @bostonglobe.com : www.bostonglobe.com/2026/04/22/o...
Good lord.
CDT’s Kristin Woelfel: “The recent determination that four school districts in Kansas violated the country’s principal federal student privacy law lacks evidence and paints a blurry picture of what the Department of Education thinks FERPA requires.” www.ed.gov/about/news/p...
Copyright is but one piece in a national framework for ensuring the security, trustworthiness, and reliability of embedded software, and other copyright-protected technology that affects our daily lives. Issues such as these extend beyond the reach of 1201 and may require a broader solution, as noted by the NTIA. The Librarian fully supports the Register in her examination of these issues and urges Congress to work with the Copyright Office and other federal agencies to consider these issues beyond the contours of this 1201 rulemaking.
Adding that the Librarian actually agreed with NTIA in 2024 on the need for reform. Unfortunately she was fired by the current Administration...
Caveat- I no longer work for NTIA, and do not know their current policy position on this issue. These are just my own thoughts based off of working with my colleagues who were leading this fight (including the only government employee to have participated in all 26 years of the process!).
Part of why NTIA was given this role by Congress was to give voice to the policy equities that lie beyond copyright infringement. Those equities have been ill-served by this process, not through fault of the Copyright Office, but due to the statute itself. It is long past time for reform.
We also remain concerned that the process can be inefficient and problematic for petitioners seeking exemptions that have been previously granted. INTIA would welcome the opportunity to discuss possible further enhancements to the rulemaking process once this proceeding is complete.
The 2020 and 2016 letters also call for reforms to the process. See below from 2016.
Limitations of the Rulemaking Process Upon reflecting on the evolution of the rulemaking and the role of technology in society since the first triennial rulemaking in 2000, we conclude that the rulemaking process itself may no longer be adequately tailored to balance the interests of rights holders and people seeking to make non-infringing uses of copyrighted works who are adversely affected by the prohibition against circumvention in Section 1201.
In 2024, NTIA's letter basically says that the rulemaking process doesn't work anymore.
This is fantastic! I would also add that NTIA has a statutory role in this process (advising the Copyright office on its recommendations), and has been publicly calling for reform for at least a decade.
I created a bombastically detailed visualization of every single comment filed in the eight rounds of these triennial rulemaking proceedings.
Five foot long posters are available for those of you with a love for elaborate governmental ritual.
1201.nyuengelberg.org/burdenofproof/
Big privacy news - over a dozen state AGs pushing for Congress to close the data broker loophole!
But I have to ask: While we're pushing Congress to act, what are these AGs doing to ban data purchases by police in their own states?
If the FCC really wanted to do something to improve router security, it would ask Congress to pass our model end-of-life disclosure act, which requires companies to disclose how long they support their connected devices, and then proactively notify consumers when those devices actually reach EOL.
Show someone who absolutely rocks a hat (that isn’t Indiana Jones)
FCC Chair Brendan Carr is weaponizing the “public interest” standard against viewpoints he dislikes. Such jawboning violates the First Amendment, due process, and the rule of law. 75+ civil society groups & free speech experts joined us in this coalition letter 🧵
techfreedom.org/fcc-threats-...
The new White House AI legislative framework is confused about the First Amendment: Gov't can't “prevent[] AI systems from being used to silence or censor lawful political expression” because AI providers’ have a First Amendment right to decide which questions to answer and how to answer them
Great article.
Let's talk about Afroman and Section 230. 🧵
Fantastic contribution/intervention from AI policy GOATs. Instant read and recommend.
Distinguishing Task-Specific and General-Purpose AI in Regulation by Jennifer Wang, Andrew Selbst, Solon Barocas, Suresh Venkatsubramanian
The task-specificity of these systems has three important impli- cations for evaluations. First, evaluations of task-specific systems have self-evident targets. ... Second, evaluations can be properly scoped. ... This allows evaluators to assess model performance, knowing that the model will only operate within this defined input space and pur- pose. Third, evaluations can be reproducible. ... These three properties allow policymakers to establish certain expecta- tions around the validity and reliability of evaluations. Although there remain technical and institutional challenges in evaluations for task-specific systems, the expectations are that (i) evaluations should appropriately characterize the expected behavior of the sys- tems relative to its defined objective (targeted), (ii) they should achieve sufficient coverage of the system’s intended uses (well- scoped), and (iii) the evaluation results should be consistent and repeatable across different independent evaluators (reproducible). The absence of task specificity in GPAI systems means that these properties fall away. GPAI systems are not trained for specific tasks but rather with broader objectives to learn structural patterns in the data (e.g. predicting the next word given the previous words). As such, there is no clear target against which to evaluate their performance. Evaluators instead target general capabilities or properties, such as linguistic understanding, reasoning , and biases. These capabilities or properties are characterized by the model’s responses to a set of custom-built and annotated scenarios, typically in the form of multiple-choice or binary questions. ... While past regulatory requirements around AI audits and impact assessments have hinged on the assumption that developers can meaningfully evaluate AI systems prior to deployment, GPAI challenges the reliability of current evaluation methods as regulatory tools.
more folks working on 'ai policy' — specifically policies seeking to prevent discrimination by ai systems in part through the use of impact assessments and audits — should read this paper by jennifer wang, @aselbst.bsky.social, @s010n.bsky.social, and @geomblog.bsky.social.
arxiv.org/pdf/2506.17347
I snuck into a concert with the choir and had to pretend to sing in a language I don’t speak by mouthing “we all live in a yellow submarine” repeatedly.
Solidarity to the NYU faculty members as they prepare to strike.
University leadership must offer a deal that meets the moment: Stronger job security, higher wages, and academic freedom protections.
NYU, it’s time to give CFU-UAW a fair contract and also pay your RAs!
The Virginia legislature has banned the sale of precise geolocation data through a simple amendment to their otherwise bad privacy law!!!
lis.virginia.gov/bill-details...
it's an extraordinarily commonsense move that is unreasonable not to adopt nationwide -- the bar is low but gotta take the W!!
Tech Talks with Jamal Magby.
On the new episode of CDT’s Tech Talks, Kate Ruane & Riana Pfefferkorn unpack the recent Grok AI controversy, from reports of harmful outputs, incl. non-consensual intimate imagery, to broader questions ab safety guardrails, moderation choices, & accountability. cdt.org/insights/tal...
Senators Mike Lee and Dick Durbin just introducing the SAFE Act, legislation to enact major reforms to FISA surveillance, in particular Section 702, a warrantless surveillance authority that’s repeatedly been misused against Americans
Here are key details about this bill….
The Verge saying the US desperately needs a #privacy law. I agree.
www.theverge.com/column/88251...
Spring Fling.
We’re counting down to CDT’s Spring Fling—where the tech policy community comes together during IAPP’s Global Summit! 🎉
📍 Crimson, DC
📅 March 30 | 7–10 PM
🎟️ Tickets & sponsorships now available: springfling@cdt.org
cdt.org/event/2026-s...
CDT's Non-Resident Fellows.
We’re thrilled to welcome CDT’s 2026–2027 class of Non-Resident Fellows — a remarkable group of scholars and experts whose insights will strengthen our work defending civil rights, civil liberties, and democratic values in the digital age. cdt.org/about/fellow...
