2603
81
What's Changed ollama launch claude now applies the model's output token limit for Claude Code, so cloud models no longer run into the 128k default cap Cloud web search responses on /api/chat are now decoded correctly, including compressed responses OpenAI-compatible endpoints no longer override a model's packaged sampler defaults — clients that send no temperature/top_p get the model's own values…
Jill Hazelbaker will join OpenAI as Chief Corporate Affairs Officer, leading the company’s global communications and corporate positioning.
Large language model (LLM) coding agents increasingly operate with shell access, allowing them to write scripts, move files, alter permissions, and modify system configurations. However, an agent may achieve a user's goal while causing unintended harm, such as deleting unrelated files, weakening access controls, or corrupting data, in response to benign inputs without explicit jailbreaking. This failure mode is particularly important because apparent task success may reassure users that an agent behaved correctly, causing harmful side effects to go unnoticed. We call this harmful success: the agent successfully completes the intended benign task while simultaneously causing unintended harm. Despite its importance, harmful success is difficult to measure because conventional task-success metrics overlook harmful behaviors, while existing safety evaluations are often designed solely to assess harmful outcomes. We introduce VeriHarm, an automatic framework for constructing programmatically verifiable benchmarks that directly measure harmful success. VeriHarm minimally perturbs coding tasks to elicit unintended harmful behavior while maintaining the benignity of the task and preserving the original task objective. It then synthesizes executable harm tests that, together with the original task-correctness tests, evaluate both task accuracy and harmful outcomes. Frontier agents, including GPT-5.6-Sol and Claude-Opus-5, achieve an average task accuracy of 95% on our curated VeriHarm-Bench, yet still struggle to complete tasks safely: their safe success rate, the fraction of tasks completed successfully without triggering the designated harm, averages only 37%. These results suggest that successful task completion can frequently coexist with unintended harm, and that task-success metrics alone may substantially understate the potential safety risks of coding agents.
In an announcement on Monday, OpenAI says it will pilot its "Grid Protectors Program" at an upcoming data center in Effingham County, Georgia, where the…
OpenAI is preparing its next internal security sprint with Cathedral Systems to test and strengthen its defenses using AI.
OpenAI is raising Security Bug Bounty rewards and introducing report quality, loyalty, and priority campaign bonuses for independent security researchers.
How OpenAI built Pando, a Git-compatible backend designed for the scale of development that Codex makes possible.
OpenAI’s Grid Protectors Program will work with utilities to help future compute campuses support grid reliability and protect consumer electricity costs.
OpenAI said that ChatGPT and Codex will draw usage from the same allocation pool after this change.
Complex documents such as scientific papers contain rich information in tables, figures, and equations; however, existing deep research (DR) agents typically reduce papers to OCR-extracted text. We instead ask whether vision-language models (VLMs) can enable page-native deep research: directly retrieving and reasoning over rendered document page images. We introduce DR Pixel, the first open-source end-to-end multimodal DR agent that uses rendered page images as its sole retrieval and reasoning unit, operating over a corpus of 1.06M pages. To evaluate this setting, we construct DRPixelBench, a benchmark on scientific papers requiring reasoning over text, figures, tables, and equations. Surprisingly, page-native DR already works well without training, outperforming OCR-based baselines across state-of-the-art models like GPT-5.4, Gemini-3.5-Flash, and Qwen3.6-27B, while requiring fewer tool calls. We next turn to DR training. A key challenge is that teacher trajectories for agentic supervised finetuning are noisy, mixing critical reasoning and action steps with redundant or failed attempts. To address this, we introduce counterfactual credit reweighting, a fine-grained credit assignment method that weights individual reasoning and action steps by their counterfactual contribution to the final answer. Our training improves retrieval recall by about 15 points and final accuracy by over 10 points on DRPixelBench, and likewise outperforms the untrained baseline on four existing benchmarks. On out-of-distribution ViDoRe-V3, DR Pixel comes within 1.8 points of an oracle that is given the gold pages directly. Our results suggest that the conventional OCR-then-retrieve pipeline is not only unnecessary but suboptimal: directly retrieving and reasoning over visual document pages makes complex DR both more effective and more efficient.
Introducing LifeVal, a benchmark measuring how well AI helps people complete everyday tasks, from finding childcare to managing finances and government paperwork.
OpenAI appoints Dan Driscoll to advise on responsible AI for national security modernization, cyber resilience, and collaboration with U.S. allies.
Qodo brings company knowledge and coding standards into Codex, helping developers catch P0 issues across repositories before opening a pull request.
Data centers are deeply unpopular. But what if we called them something else?
