Andy Jung

AI Policy Counsel at @TechFreedom. Tech policy professional: AI, admin law & liberty. Also Boston Terriers. DC native in the LBC. No legal advice.

[Andy Jung @AndyJungTech]

New piece from me on laws banning minors from using chatbots.

Minors have a First Amendment right to receive information, and the vast majority of chatbot outputs are protected speech. Full-on bans go beyond protecting minors from obscene material—they're not narrowly tailored.

[Andy Jung @AndyJungTech]

@deanwball I get some good mom support as well

[Andy Jung @AndyJungTech]

@devahaz I'll stop spamming you but this is too good

https://t.co/eCyToDPTdu

[carl1tosalcaraz @carl1tosalcaraz]

me anytime I see zver*v winning

[Andy Jung @AndyJungTech]

@devahaz Vamos Fonseca. Him and Felix are the only two I can realistically see challenging Zverev. Although I should probably just keep my mouth shut and avoid the jinx.

[Andy Jung @AndyJungTech]

@devahaz This aged terribly.

[Andy Jung @AndyJungTech]

@devahaz Really hoping Jodar shows us something.

[Andy Jung @AndyJungTech]

@Ashkhen @TechFreedom @corbinkbarthold MRW Congress introduces another age verification AI bill:

[Andy Jung @AndyJungTech]

Good rebuttal to the voluntary point, although the federal government could have accomplished this—create a Secure Frontier Model Deployment benchmarking process for deploying models and require it for gov contracts—without an EO.

[Jessica Tillipman @JTillipman]

Initial reaction to the new AI EO:

No company is formally required to participate, but if a developer wants to sell frontier AI systems to the federal government, participation may soon become the price of entry.

Agencies can build engagement with the process into solicitations, evaluation criteria, and cybersecurity requirements.

So yes, the process is "voluntary," but "voluntary" looks different when the federal government is a dominant customer and refusal can result in procurement consequences.

[Andy Jung @AndyJungTech]

@Ashkhen @TechFreedom @corbinkbarthold You hired and mentored me as an intern. Thanks Ash-I owe you a lot and am grateful!

[Andy Jung @AndyJungTech]

@nkrosse @TechFreedom Thanks Nick!

[Andy Jung @AndyJungTech]

@CharlieBull0ck Agreed

https://t.co/umvJZNJZCU

[Andy Jung @AndyJungTech]

The Promoting Advanced AI Innovation and Security EO seems net fine to me.

Strengthening infrastructure and prioritizing cyber readiness are plainly good.

And the Secure Frontier Model Deployment benchmarking framework is, after all, voluntary.

I do worry that the federal government will retaliate against labs that choose to release models without going through the process / giving the fed a heads up. See Anthropic debacle (the federal government retaliating against an AI lab for refusing its terms). I also understand the concern over classifying the benchmarking framework, but the federal government already has processes and rules for creating unclassified versions of classified frameworks to share with relevant parties without security clearance. I think (or at least hope) AI developers and researchers will get their hands on docs containing the high-level criteria and threshold definitions of the framework to inform their work.

So, some good - some meh = net fine.

[Andy Jung @AndyJungTech]

Very stoked about this. I plan to expand the scope of my AI work to cover AI governance more broadly, and I'm especially excited to focus on the inevitable wave of AI-related litigation which is forming on the horizon. Let's collaborate on cool projects.

[TechFreedom @TechFreedom]

🎉 We are excited to announce the promotion of @AndyJungTech to our first AI Policy Counsel!

Andy has played a key role in advancing our work on emerging technology policy, and this promotion reflects both his contributions and the growing importance of AI policy to our mission.

[Andy Jung @AndyJungTech]

The Promoting Advanced AI Innovation and Security EO seems net fine to me.

Strengthening infrastructure and prioritizing cyber readiness are plainly good.

And the Secure Frontier Model Deployment benchmarking framework is, after all, voluntary.

I do worry that the federal government will retaliate against labs that choose to release models without going through the process / giving the fed a heads up. See Anthropic debacle (the federal government retaliating against an AI lab for refusing its terms). I also understand the concern over classifying the benchmarking framework, but the federal government already has processes and rules for creating unclassified versions of classified frameworks to share with relevant parties without security clearance. I think (or at least hope) AI developers and researchers will get their hands on docs containing the high-level criteria and threshold definitions of the framework to inform their work.

So, some good - some meh = net fine.

[Andy Jung @AndyJungTech]

Claude, convince my landlord to start accepting rent payments via Zelle. Remind him it's the big 2026. Make no mistakes.

[Andy Jung @AndyJungTech]

Has never even occurred to me to do this, I'm proud to say.

[Séb Krier @sebkrier]

Asking an LLM which fictional character you are is astrology for boys. :)

[Andy Jung @AndyJungTech]

Day 1 of my road trip to Big Sur. Stopped for the night in Cambria

[Andy Jung @AndyJungTech]

@deanwball Great takes all-around

[Andy Jung @AndyJungTech]

Waffles

[Big baby🌸 @lizzyjollof]

Pick a bed

[Andy Jung @AndyJungTech]

@anton_d_leicht A "light touch" EO wouldn't have precluded political, inscrutable, off-record conversations either. Especially under this administration

[Andy Jung @AndyJungTech]

@devahaz @jbarro Right...and that's embarrassing. All learning involves some trial and error. Reinforcing that point isn't ableist, it's just life

[Andy Jung @AndyJungTech]

@FTC @rytr_me @AFergusonFTC The proposed rule only mentioned artificial intelligence once, deep in a footnote. But it clearly would have applied to AI as well as software like Photoshop.

[Andy Jung @AndyJungTech]

New @FTC AI-ish related enforcement action. Interesting because (1) it falls into the growing bucket of deception actions against companies that falsely claim to use AI, and (2) the use of "means and instrumentalities" liability 🧵

[FTC @FTC]

FTC to require Cox Media Group, two other firms to pay nearly $1 million to settle charges they deceived customers about “active listening” AI-powered marketing service: https://t.co/i8MDDkLbBH

[Andy Jung @AndyJungTech]

@FTC @rytr_me @AFergusonFTC The proposed rule did not define “means and instrumentalities” of impersonation, saying only: “One who places in the hands of another a means or instrumentalities to be used by another to deceive the public in violation of the FTC Act is directly liable for violating the Act.”