The strategic choice of model "openness" has become a defining issue for the foundation model (FM) ecosystem. While this choice is intensely debated, its underlying economic drivers remain underexplored. We construct a two-period game-theoretic model to analyze how openness shapes competition in an AI value chain, featuring an incumbent developer, a downstream deployer, and an entrant developer. Openness exerts a dual effect: it amplifies knowledge spillovers to the entrant, but it also enhances the incumbent's advantage through a "data flywheel effect," whereby greater user engagement today further lowers the deployer's future fine-tuning cost. Our analysis reveals that the incumbent's optimal first-period openness is surprisingly non-monotonic in the strength of the data flywheel effect. When the data flywheel effect is either weak or very strong, the incumbent prefers a higher level of openness; however, for an intermediate range, it strategically restricts openness to impair the entrant's learning. This dynamic gives rise to an "openness trap," a critical policy paradox where transparency mandates can backfire by removing firms' strategic flexibility, reducing investment, and lowering welfare. We extend the model to show that other common interventions can be similarly ineffective. Vertical integration, for instance, only benefits the ecosystem when the data flywheel effect is strong enough to overcome the loss of a potentially more efficient competitor. Likewise, government subsidies intended to spur adoption can be captured entirely by the incumbent through strategic price and openness adjustments, leaving the rest of the value chain worse off. By modeling the developer's strategic response to competitive and regulatory pressures, we provide a robust framework for analyzing competition and designing effective policy in the complex and rapidly evolving FM ecosystem.

The developments in Generative AI technologies have paved the way for numerous innovations in different fields. Recently, Generative AI has been proposed as a competitor to AES systems in evaluating student essays automatically. Considering the potential limitations of AI in processing idioms, this study assessed the scoring performances of Generative AI models for essays with and without idioms by incorporating insights from Corpus Linguistics and Computational Linguistics. Two equal essay lists were created from 348 student essays taken from a corpus: one with multiple idioms present in each essay and another with no idioms in essays. Three Generative AI models (ChatGPT, Gemini, and Deepseek) were asked to score all essays in both lists three times, using the same rubric used by human raters in assigning essay scores. The results revealed excellent consistency for all models, but Gemini outperformed its competitors in interrater reliability with human raters. There was also no detectable bias for any demographic group in AI assessment. For essays with multiple idioms, Gemini followed a the most similar pattern to human raters. While the models in the study demonstrated potential for a hybrid approach, Gemini was the best candidate for the task due to its ability to handle figurative language and showed promise for handling essay-scoring tasks alone in the future.

Monte Carlo (TSMC) to provide scalable stepwise supervision using small LMs. This yields more efficient training, improved stability, and higher accuracy. OpenAI's o1 series (OpenAI, 2024) introduce inference-time scaling by increasing the length of the Chain-of-Thought (CoT) (Wei et al., 2022) reasoning process. Despite their empirical success, RL approaches that generate the entire reasoning chain in a single forward pass face notable limitations, including CoT derailment, where the reasoning trajectory drifts off course due to accumulated errors, and the inherent challenges of long-horizon RL with sparse outcome rewards. This sequential scaling strategy, i.e., simply extending the CoT length, can therefore be insufficient (Y ang et al., 2025). To improve planning quality, we introduce Multi-Path Plan Aggregation (MPP A). For each planning step, the model generates multiple alternative plans and aggregates them into an improved plan before proceeding to the subsequent execution steps. Beyond enhancing planning, we identify a fundamental challenge in credit assignment for long-horizon policy learning (Kaelbling et al., 1996). Existing RL fine-tuning frameworks struggle to provide effective process-level supervision (Guo et al., 2025). First, evaluating the correctness of intermediate steps is inherently difficult. Automated annotation using LLM judges (Gu et al., 2024) often yield unreliable or noisy signals Second, introducing a separate process reward model (PRM) adds complexity. We then define the process preference between two candidate continuations at the same step by comparing their incremental log-weights. We repurpose Twisted Sequential Monte Carlo (TSMC) to provide process-level preferences for online Step-DPO training. Results show that our approach consistently outperforms both distillation-based long-CoT methods and RL methods that rely solely on outcome rewards. The Chain-of-Thought trajectories can be lengthy and the positions of the first error vary considerably, making outcome-based RL fine-tuning inefficient. Training long trajectories with outcome rewards is highly inefficient.

OpenAI patched the ShadowLeak attack in August 2025 after hackers exploited ChatGPT to exfiltrate Gmail data through invisible email prompts without user interaction.

An illustration of how AI manipulates and changes images. An illustration of how AI manipulates and changes images. Ask Google's AI video tool to create a film of a time-travelling doctor who flies around in a blue British phone booth and the result, unsurprisingly, resembles Doctor Who . And if you ask OpenAI's technology to do the same, a similar thing happens. What's wrong with that, you may think?

The precarious work of training AI, which generally pays just a few dollars, has sparked a movement for better wages and conditions globally. Paris - For a generative artificial intelligence system to learn how to write an autopsy report, human workers must sort and annotate thousands of crime scene images. The precarious work of training AI, which generally pays just a few dollars, has sparked a movement for better wages and conditions stretching from Kenya to Colombia. You have to spend your whole day looking at dead bodies and crime scenes. In a time of both misinformation and too much information, quality journalism is more crucial than ever. By subscribing, you can help us get the story right.

OpenAI has temporarily stopped its artificial intelligence (AI) app Sora creating deepfake videos portraying Dr Martin Luther King Jr, following a request from his estate. It said disrespectful content had been generated about the civil rights campaigner. Sora has become popular in the US for making hyper-realistic AI-generated videos, which has led to people sharing clips of deceased celebrities and historical figures in outlandish and often offensive scenarios. OpenAI said it would pause images of Dr King as it strengthens guardrails for historical figures - but it continues to allow people to make clips of others. The firm has faced controversy over this stance, as videos featuring notable figures such as President John F. Kennedy, Queen Elizabeth II and Professor Stephen Hawking have been shared widely online.

In this era of AI slop, the idea that generative AI tools like Midjourney and Runway could be used to make art can seem absurd. But amid all the muck, there are people using AI tools with real consideration and intent. Some of them are finding notable success as AI artists: They are gaining huge online followings, selling their work at auction, and even having it exhibited in galleries and museums. This story is from our forthcoming print issue, which is all about the body. Plus, you'll also receive a free digital report on nuclear power. Take our quiz: How much do you know about antimicrobial resistance?

After launching in October in the US and Canada via invitation only, OpenAI's video app, Sora 2, hit 1m downloads in just five days. After launching in October in the US and Canada via invitation only, OpenAI's video app, Sora 2, hit 1m downloads in just five days. The video app can produce realistic deepfakes of Marx shopping and MLK Jr trolling. Some say using'historical figures' is the company's way of testing the legal waters L ast night I was flicking through a dating app. One guy stood out: "Henry VIII, 34, King of England, nonmonogamy".

Many will assume that OpenAI's Sora app represents a new era of social media. But that's wrong--all it does is reanimate our current one. As a purely creative instrument, Sora, the new AI video app from OpenAI, is a game changer. Dream up any scenario and it appears in an instant. Mr. Rogers teaching Tupac Shakur the lyrics to the legendary rap diss "Hit Em Up."