/Tech4h ago

Box CEO Aaron Levie argues model retractions incentivize sovereign open weights, while Ethan Mollick questions their financial viability

Story Overview

Days after the US ordered Anthropic to cut off access to its newest frontier models for foreign users, the abrupt global retraction has sparked fresh debate: Levie sees it as a catalyst that could push countries toward sovereign open-weights systems to avoid depending on US-controlled tech, while Mollick counters that the economics simply do not support training competitive frontier open models at scale.

64383389635.2K

#184

Original post

Aaron Levie@levie#195inTech

The big winner in all of this is going to be open weights models. This is a huge win for the field, as a risk that was entirely theoretical and untested 2 days ago (that a model could be pulled back), now has a new precedent that’s been set.

The game theory the US should highly consider, and the risk with regulating AI at the model layer vs. applied layer, is that other countries now have even more incentive to develop sovereign AI.

If at any moment a model can be become unavailable to your country’s users or businesses, this poses very real risk on relying on technology from a particular country.

As a result, it forces major countries to charter their own path on AI development, which reduces America’s leadership role in this tech stack over time. The most likely solution that other countries will rely on is open weights models, which currently is generally not coming from the US.

America should be considering all of these downstream implications as it decides how and where in the stack to be regulating AI. At the same time, we should be doing a ton more OSS innovation.

7:35 AM · Jun 14, 2026 · 34.7K Views

Policy Risk

Layering rules at the app level might dodge the sovereign scramble

Levie argues that targeting the model layer itself creates exactly the kind of precedent that encourages other nations to develop independent alternatives, risking a fragmented landscape where US leadership slips.

Open Question

The cash gap keeps frontier open weights on the sidelines

Mollick notes that while smaller open models can still find a path, no clear financial incentives exist to train the largest open-weights systems, leaving a persistent capability divide between closed and open approaches.

Sentiment

Users backed open weights models as safer after AI regulations risked turning hosted access into a government lever prone to sudden cuts like the Fable incident.

Pos

94.7%

Neg

5.3%

20 comments with sentiment.

Cluster Engagement

Posts from X

Most Activity

VIEWS3.1KBOOKMARKS4LIKES59RETWEETS1REPLIES7

Ethan Mollick@emollick

@levie The financial incentives for training frontier open weights models don’t exist. Small models, sure, but there is a huge capability gap.

Aaron Levie@levie

The game theory the US should highly consider, and the risk with regulating AI at the model layer vs. applied layer, is that other countries now have even more incentive to develop sovereign AI.

If at any moment a model can be become unavailable to your country’s users or businesses, this poses very real risk on relying on technology from a particular country.

America should be considering all of these downstream implications as it decides how and where in the stack to be regulating AI. At the same time, we should be doing a ton more OSS innovation.

2h3.1K594

Marcus Mendez@Marcus4cyb

@emollick @levie yeah but the costs fall .. the chinese do have a financial incentive to build their own industry .. monetize it at other layers ( like industrial semi layer , robots , cars )

1h15

ilya@ibaklanov

@levie Do you think USG could ban open weight models like they ban various illegal information (how to make drugs, weapon blueprints, etc)?

3h378

Bella Casa@mxworxtonic

@emollick @levie Financial incentives for training frontier closed weights models also do not exist, Except for - those whose main business is under threat. - and who have already sunk billions in this race.

2h411

Greg@gentschev

@levie Considering downstream consequences? In this administration?

3h1653

Marcus Mendez@Marcus4cyb

@emollick @levie keep in mind , mythos/fable is like what, 6mo behind where they are if not more. the cost then would have fallen to start from scratch now.. what level do you need to be disruptive? exactly why china blew the world up in solar and evs not that they really could get most benefits

1h15

Marcus Mendez@Marcus4cyb

@emollick @levie china is starting to look really really good here.. solar panels, evs , AI Models, humanoids

1h12

LeetLLM.com@leetllm

@levie The real story here isn't just open weights winning, it's that the API kill-switch is no longer a theoretical risk. Fable 5 vanishing overnight is the ultimate ad for Qwen3.6 and DeepSeek V4 Pro.

2h21

Dee@dee_hw

@levie Good take. After this Fable cut off, every country must have an AI strategy now.

4h2071

NCD@Jokerbernrin

@levie Open weights remove a lot of uncertainty

4h1671

Julian Goldstahl ack/me@JulianGoldstahl

@levie Feeble Fable.

4h1031

Russell@russell_m

@levie I hope you are right, but I think it is a little naive to think a hostile (to AI) government could not block Open Models. It could turn into 1920s prohibition all over again.

4h317

Joe Johnson@joerjohnson

@levie Agree, and model regulation entrenches the few Frontier labs that are big enough to clear a review board.

2h941

Andreas Ramos@Andreas_Ramos

@levie what will happen when (when, not if) China develops a smiliar AI? the US bans its use? and Europe and India? it will happen.

3h282

what do you think of this one vega?@s_pect_re

@emollick @levie businesses might need financial incentives, but nations don't.

2h262

Nav@thakur_nave

@levie The models were already regulated prior to friday. More people are learning about this now.

3h225

Diam@diamai_

@levie Open weights solve the off-switch problem. I’m less sure they solve the speed problem. Can that ecosystem keep up with closed labs burning insane money on compute every month?

3h173

Jatin Garg@jatingargiitk

this is the right call and worth pushing further. regulating at the model layer turns model access into a lever, and every lever a government holds is a reason another country builds its own. you can't be both the world's ai supplier and the country that can cut supply. pick one, because everyone else is now planning around the second.

3h156

sidharth satapathy@satapathy9

@levie The opportunity to innovate on feedback is gone now. Bias will make it even more worse as it evolves further. From a true technical standpoint this is the worst this a country could do to bright minds solving bigger problems for mankind.

4h153

Peter Dedene@dedene

@levie @LyraInTheFlesh Closed models are now a proven single point of failure. Most enterprises will not tolerate this.

3h401