Mistral Large 4 is available for an API preview, while its weights are planned for release later this month. Openness depends not only on launch language, but on when users can inspect, deploy, and independently evaluate a model.
AI
Public thoughts on artificial intelligence, creativity, and human-machine collaboration.
At a New York City Council hearing, former employees warned of AI risks while company representatives stressed benefits and safety. Participants still disagreed on how to quantify catastrophic risk and who is responsible if harm occurs.
A voluntary White House agreement with AI companies calls for internal controls and external audits, but leaves enforcement details and consequences unclear. A former OpenAI safety lead has also shared a personal account, leaving the gap between promises and accountability worth watching.
Reports say the White House is expected to name the current director of national intelligence as its AI adviser, but no formal appointment has been announced and the role remains undefined. The choice signals a security lens; its institutional effects remain to be seen.
A new report says automated AI R&D could create a feedback loop, while its authors also describe the evidence as preliminary and the outcome as uncertain. It matters to distinguish a possibility from an established fact.
A new report co-authored by more than 20 researchers asks whether AI taking part in its own R&D could create a feedback loop, placing possible risks, current evidence and policy proposals in the same frame.
A proposed AI data campus still at the planning stage touches land, energy and water in a biosphere reserve, making public access to the full plan part of the debate.
Historical texts from Oxford’s Bodleian Library are being used to train models, potentially widening access while raising questions about use, disclosure and the responsibilities of public institutions.
AI does not live only in product launches and capability rankings. Comedy that takes apart technology slogans is also a reminder that the future need not arrive exactly as advertising says it must.
AI is changing how people reach sources and check claims. Convenience is real, but the responsibility to verify has not gone away.
As frontier models become cheaper, the question is whether safety evidence, independent assessment, and public accountability can keep pace.
AI reporting is moving from what systems can do toward the evidence and standards that different countries and labs should share.
As AI becomes a national capability, the question is not only what models can do but who sets the pace, receives the benefits, and bears the risks.
As AI capability moves closer to real systems and public decisions, the responsibility to prove safety and reliability must arrive before the consequences do.
As models enter research, testing, and attack chains, safety cannot wait outside the product release. Evaluation, permissions, logs, and shutdowns need to become work that can be checked every day.
Recent reports place the AI race, ballot privacy, corporate self-regulation, and the pace of oversight on the same table. The question is not who leads first, but who can make the risks clear before leading.
Recent reports place the AI race, ballot privacy, corporate self-regulation, and the pace of oversight on the same table. The question is not who leads first, but who can make the risks clear before leading.
Recent reports place the AI race, ballot privacy, corporate self-regulation, and the pace of oversight on the same table. The question is not who leads first, but who can make the risks clear before leading.
From autonomous drone software and malicious code to calls from industry leaders to slow development, recent reports place AI capability and consequence on the same table.
Recent reports place AI alongside weapons, regulation, and children’s education. The question is not only what it can do, but when it must stop.