YouCongress - Expert and Citizen Preferences

Releasing full model weights may allow malicious actors to strip or override safety mechanisms, creating uncensored or harmful versions. In contrast, supervised fine-tuning preserves core safety guardrails while enabling responsible customization.

Restricting access to a dual-use AI model is precautionary only if it delays harmful actors more than defenders. That condition varies across actors: a state agency or organized criminal group may obtain a substitute through theft, distillation, inte...

Like policymakers, we want a safe, secure AI ecosystem, and view openness as an important part of achieving it. We support policies that preserve the ability to develop, distribute, and use open weight models subject to focused, risk-based safeguards...

Think of an AI agent that starts issuing refunds or firing off API calls in a way you did not intend. A kill switch wrapped around that behavior lets you disable it in one action while you investigate, instead of scrambling for an incident bridge.

The kill switch was the right metaphor for a moment that is already passing. The dimmer switch is closer to how regulated enterprises actually need to govern agents that are now part of how the work gets done.

In selecting AI products to use internally, we will prefer tools that allow us to opt out of having our data used in the product’s training data.

We are glad the Government agrees that creative workers should be paid fairly for their work. We have campaigned tirelessly to ensure that the unworkable opt-out and a commercial Text and Data Mining (TDM) exception, which would enable unrestricted u...

Devin's appointment gives us the operational and technical leadership to match the scale of what we're building, and the Frontier Security Institute extends our mission into territory that has never been more consequential. Frontier AI is now a natio...

This Article argues that AI threatens to disrupt the tax system’s ability to fulfill its fundamental goals of raising revenue, redistributing income, and regulating taxpayer behavior. The risk is heightened by the federal government’s dependence on i...

Raising the cost of all those innovation- and productivity-enhancing uses of computing would unequivocally harm American productivity, innovation, and global competitiveness. Knowingly imposing these costs on ourselves because of concerns about certa...

The Civic Assembly in Snohomish County thus provides an example of how citizen participation through random selection can also be used in relation to future technological issues: not as a substitute for parliamentary decisions, but as a well-founded ...

A frontier developer shall report any critical safety incident pertaining to one or more of its frontier models to the office within seventy-two hours from a determination that a critical safety incident has occurred or within seventy-two hours of th...

Timelines for incident reports are often set with well-intentioned urgency, but if they implicitly assume that root-cause analysis or full impact assessment has already been conducted, they will either produce premature and unreliable reports or plac...

When data theft occurs inside 60 minutes, a 72-hour reporting window ceases to function as an effective defense mechanism. Instead, it becomes a post-mortem.

Congress is moving to require clear disclosures whenever political advertisements use AI-generated content. Representative Yvette Clarke introduced legislation mandating that voters know when AI has been used to create or modify political messages th...

We should ask ourselves an important question: does the problem require government regulation?

In the coming years, AI — and ultimately artificial general intelligence (AGI) — has the potential to drive one of the greatest transformations in history. We’re a team of scientists, engineers, ethicists and more, working to build the next generatio...

This is going to be the most productive decade in the history of our species. But in order to truly reap the benefits of AI, we need to learn how to contain it. Paradoxically, part of that will mean collectively saying no to certain forms of progress...

We argue that general intelligence may be the only biological strategy that generates existential threats to the species possessing it, an existential risk paradox with no parallel among non-intelligent survival strategies. Unlike prevailing accounts...

A compute permit system would face its own challenges: defining which training runs qualify as high-stakes, preventing developers from fragmenting workloads to stay below thresholds, and ensuring that monitoring infrastructure does not become a tool ...

According to the proposed rule, a covered US person will be required to provide quarterly reports to BIS if the US person “engages in, or plans, within six months, to engage in ‘applicable activities.’” “Applicable activities” include: “Conducting a...

The current thresholds (10^25 FLOP in the EU AI Act and 10^26 FLOP in the now rescinded US Executive Order) may not be the most effective risk indicators.

Equity funding: policymakers should expand the mechanisms by which these accounts can be funded, including with equity in AI companies, so that beneficiaries share directly in nearer-term gains from AI-driven growth.

There’s something very interesting about it, where it almost becomes a partnership with the American public.

It puts the government in the space where it’s no longer focused on ensuring the U.S. has the capacity it needs to protect the public interest and is more focused on ensuring that its investment pays off.

All else equal, prefer architectures that require CoTs for serial reasoning, until interpretability methods can provide similar levels of insight.

We have a very natural instinct to say: we want as much interpretability as possible. One of the sci-fi worry cases is that agents start speaking in languages to each other that we don't understand. [...] I bet there are ways to define boundaries tha...

[...] principles requiring that AI be explicable are misguided. We should be deciding which decisions require explanations. Automation is still an option;

An upcoming SSDF update should mandate automated, independent verification layers because AI models cannot reliably self-certify their own probabilistic outputs.

SIIA raised particular concerns about applying traditional product liability frameworks to generative AI, regulating frontier models at the state level, and imposing premature audit and transparency requirements without established standards.

We can’t just invest in software and hardware; we have to invest in sovereignty.

Europe should prioritize “code over concrete.”

As these tools become part of everyday life, it's important that they're designed to be safe, trustworthy, and appropriate for different stages of development. That's why independent evaluation and public accountability matter.

However, it is important to differentiate regulating applications (which we need) vs. regulating the technology (which is ill-advised).