The Biden White House’s February 2024 move on open and closed AI was a request for evidence—not a ban, release mandate, or finding that open models are unsafe. The National Telecommunications and Information Administration (NTIA) asked what risks and benefits could follow when powerful AI models’ weights are widely available. Its July 2024 report urged continued monitoring and evidence-gathering, while leaving open the possibility of later action if risks warranted it.
What “open” and “closed” AI mean in this debate
The formal focus was “dual-use foundation models with widely available model weights.” Weights are numerical parameters that shape a model’s outputs. They are only one part of a system: developers may also release code or training data, restrict access, or disclose some components while keeping others private.
NTIA’s inquiry centered on powerful models trained on broad data, generally through self-supervision, that can be used across contexts and can perform—or be readily modified to perform—tasks posing serious risks. Its notice described covered models as having at least tens of billions of parameters; that threshold defined the inquiry’s scope, not a finding that size alone predicts danger. NTIA also invited comment on models beyond that scope to understand the wider landscape. NTIA’s 2024 request for comment set out the scope.
“Open” and “closed” are therefore shorthand, not two exhaustive categories. As NTIA Administrator Alan Davidson put it in February 2024, “There are gradients of openness.” A fair comparison asks which components are available, who can access them, whether use is conditional, and what monitoring or accountability remains possible.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
#1 Best Overall
Why supporters argue for wider access
Supporters say access to weights can broaden participation in AI research and development, give startups and researchers more room to adapt systems, and decentralize control from a small number of providers. NTIA’s report also noted that users may run models without sending their data to a third party. The White House’s 2025 AI Action Plan made a similar case: startups could adapt open models without depending on closed providers, and businesses or governments could keep sensitive data from those vendors. The 2025 Action Plan also cited academic research and geopolitical positioning as reasons to encourage open-source and open-weight AI.
These benefits depend on the release and the user. Having weights available does not automatically make a model easy to inspect, affordable to run, or accessible to everyone. Cornell researcher David Gray Widder told the Associated Press that practical use can still require resources concentrated among large companies. Nor does openness necessarily include training data or the ability to verify how the system was built.
Rank #2
Why wider availability raises security questions
Once weights are public, developers may have less control over how others adapt or deploy the model. The Biden executive order and NTIA’s notice highlighted concerns including removal of safeguards, misuse, limited oversight, accountability gaps, security, public safety, equity, privacy, and civil rights. These are risks to evaluate—not proof that every open-weight model creates the same level of danger.
Release choices can also differ within one company’s lineup. In February 2024, Google released Gemma, open models derived from technology used for Gemini, while its more powerful Gemini offering remained closed. That example illustrates different access choices; it does not establish that one approach is categorically safer.
Recommended Free Tools
In its report, NTIA treated the risks and benefits as marginal risks to assess against other technologies and described its review as non-exhaustive and not definitive. The materials do not establish a definitive comparative rate of harm for open versus closed models. NTIA’s July 2024 report explains the limits of the evidence and its recommendation.
What the 2024 White House initiative did
On February 21, 2024, NTIA opened a 30-day public comment period as part of President Joe Biden’s October 2023 AI executive order. The formal notice set March 27, 2024, as the deadline. It asked how widely available weights and other model components might affect the economy, communities, individuals, and national security. The request sought views on potential risks, benefits, and appropriate policy approaches; it did not impose a new restriction.
NTIA said it received 332 written comments. That is a count of submissions to the consultation, not a measure of public consensus or evidence that a particular harm or benefit occurred. NTIA’s announcement of the comment count reported the number.
What NTIA concluded—and what it did not
On July 30, 2024, NTIA concluded that the evidence then available could not definitively establish either that restrictions on open weights were warranted or that restrictions would never be appropriate. It recommended monitoring a portfolio of risks and developing the ability to collect evidence, evaluate changes, and respond if heightened risks emerge.
Best Value
That was a cautious, conditional recommendation—not an endorsement of unrestricted release and not a prohibition. It kept future intervention available if evidence justified it, without claiming the evidence already settled the question.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How the later federal position differed
The White House’s July 2025 AI Action Plan called for encouraging open-source and open-weight AI, citing flexibility for startups, sensitive-data handling, academic research, and geopolitical positioning. It also said whether and how a model is released remains fundamentally the developer’s decision. This later statement shows a policy preference in that document; it does not establish that every proposed action has been implemented or resolve the technical debate.
The 2024 consultation and 2025 plan should not be collapsed into one decision: the first sought evidence and produced a conditional monitoring recommendation; the later document advocated encouraging open models.
What to compare when evaluating a model release
Rather than asking only whether a model is “open,” consider the specific trade-offs:
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problems- Available components: Are weights, code, and training data all available, or only some?
- Access terms: Can anyone use the materials, or are there eligibility rules or conditions?
- Capability and modification: How capable is the model, and how readily can it be fine-tuned or adapted?
- Oversight and accountability: Can downstream use be monitored, and who can respond when the model is misused?
- Safeguards: Can protections be removed or bypassed after release?
- Privacy and practical access: Can users run the model locally to keep data private, and do they have the computing resources to do so?
No single factor settles the policy question. Greater access can enable useful research and control over data, while also making some forms of oversight harder. The central challenge raised by NTIA was how to weigh those consequences as model capabilities and evidence change.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
