AI maker Anthropic says the US government has lifted an export ban on its powerful Mythos and Fable systems. AI maker Anthropic says the US government has lifted an export ban on its powerful Mythos and Fable systems. Anthropic has said the US commerce department has lifted export controls on its Fable and Mythos AI models, less than three weeks after the company was ordered to suspend access to its most advanced AI models over national security risks. "We'll begin restoring access tomorrow," Anthropic said in a statement on X late on Tuesday. US authorities blocked access to the models on national security grounds several weeks ago, but in a letter to Anthropic seen by Reuters, US commerce secretary Howard Lutnick, said the export controls were withdrawn and that a licence was no longer required for their export.

CIA Director John Ratcliffe speaks during a news conference in the James S. Brady Press Briefing Room at the White House in Washington on April 6. | REUTERS WASHINGTON - CIA Director John Ratcliffe on Tuesday compared the capabilities of the most advanced artificial intelligence models to nuclear weapons in a tacit defense of Washington's recent hard line on controlling the release of the most powerful AI technology. "In conversations with many of the president's other national security and economic security advisors, we're talking about the impact of these frontier AI models," Ratcliffe said during a speech at the AWS summit in Washington. "It would be ... not misplaced to refer to their capabilities as akin to digital nuclear weapons," he said. On June 12, U.S. President Donald Trump's administration forced Anthropic, a leading American AI firm based in San Francisco, to cut off access to its two most powerful models, Mythos 5 and Fable 5, by imposing an export control on them. The forced withdrawal of a frontier model by a government -- a first -- was only partially lifted on Friday for Mythos, now accessible to a restricted circle of U.S. partners, while Fable 5, its restricted consumer version, remains offline. OpenAI, Anthropic's American rival, launched its GPT-5.6 model the same day with very limited access, agreeing for the first time to let the U.S. government vet authorized partners on a client-by-client basis.

Airbnb says the "anti-party system" it deploys ahead of major holiday weekends flags bookings with characteristics indicating guests are likely to use properties for unauthorized parties.

Birds' nests of fiber-optic cables show war's impact on Ukraine Yana Hrynko, senior researcher of The National Museum of the History of Ukraine in the Second World War, shows a bird's nest made with fragments of optic fiber, which was found by a Ukrainian service member on the front line and then passed to the museum in Kyiv on June 23. Kyiv - Woven from fiber-optic cable and grass, a small bird's nest found near the front line of the war in Ukraine shows how the more than 4-year-old conflict is reshaping the natural environment, researchers say. Areas along the 1,200-kilometer front line are covered with ultrathin fiber-optic cables, which are used by Ukrainian and Russian troops to guide aerial attack drones to make them impervious to electronic jamming. The cables, which can stretch for 20 km, lie tangled in trees and scattered across fields and on the rooftops of towns in Ukraine's frontline regions, glistening in the sunlight like giant spider webs. Birds have begun repurposing the discarded cables to weave their nests, says Yana Hrynko, a senior researcher at Kyiv's War Museum, cautiously examining two delicate nests the armed forces sent to the museum from the front line. "Objects such as bird nests with fragments of optic fiber demonstrate the change in the nature of war," said Hrynko.

The US government has lifted export controls on Anthropic's most advanced artificial intelligence (AI) tools, just weeks after ordering it to restrict access to them over national security concerns, the company has said. Anthropic said in a social media post that it will begin restoring access to Claude Fable 5 and Mythos 5 on Wednesday after being notified that the US Department of Commerce has lifted export controls on the two models. They are the firm's most advanced AI tools, which were abruptly suspended on 12 June over concerns that they could be used by hackers to exploit weaknesses in computer systems. The BBC has contacted the Department of Commerce for comment. Mythos and Fable are two of Anthropic's AI models built on its Claude platform - a rival to the likes of OpenAI's ChatGPT and Google's Gemini. Fable 5 is a version of the AI model for the cosumer market, capable of deep reasoning and can perform complex tasks independently.

Deep neural networks (DNNs) frequently fail to generalize to out-of-distribution (OOD) medical images because of variations in scanners and acquisition protocols. Retraining DNN models to address these distribution shifts is often impractical due to the high cost of acquiring and annotating new medical datasets. To address this, we introduce VarDeepPCA, a novel lightweight variational DNN framework designed to restore/refine degraded segmentation maps by leveraging intrinsic geometric priors. Unlike existing approaches that require target-domain data or extensive pre-training, our VarDeepPCA explicitly learns a distribution of valid anatomical geometries using only small in-distribution (ID) datasets. Theoretically, our novel variational learning framework leverages a reinterpretation of the softmax mapping to implicitly perform exact distribution modeling, thereby enabling computationally efficient, sampling-free learning and inference. This also enables VarDeepPCA to provide uncertainty estimates associated with its restored segmentation maps. We empirically validate our framework across 4 distinct clinical applications, using 14 publicly available datasets, involving segmentation of the myocardium, neuroretinal rim, prostate, and fetal head. Comparisons against 15 existing methods demonstrate that VarDeepPCA consistently restores segmentation maps produced by the existing methods on OOD data to (i) significantly improve anatomical plausibility of geometries and clinical utility of the segmentations, and (ii) significantly reduce errors, without needing any more training data than that used by existing methods.

Self-evolution offers a scalable path to stronger reasoning: a pretrained language model improves itself with only minimal external supervision. Yet existing methods either depend on extensively curated or teacher-generated training data, or, when the generator runs unsupervised, reward it by a difficulty heuristic that need not improve the solver. We introduce INFUSER, an iterative co-training framework with two co-evolving roles: a Generator that drafts questions and reference golden answers from a pool of unstructured, automatically collected documents, and a Solver that improves by training on them. The solver is trained with standard correctness rewards against the generator-provided answers, while the generator is rewarded by an optimizer-aware influence score that measures whether each proposed question would actually improve the solver on the target distribution. Because this continuous, noisy influence score is poorly served by standard GRPO, we propose DuGRPO, a dual-normalized variant of GRPO, for generator training. Together, these turn the document pool into an adaptive curriculum that favors questions useful to the current solver, not just hard ones. On Qwen3-8B-Base, INFUSER outperforms strong self-evolution baselines with over 20% relative improvement on Olympiad and SuperGPQA benchmarks, and an 8B INFUSER co-evolving generator outperforms a frozen 32B thinking generator on math and coding. Ablations confirm each design choice is necessary, and two extensions, applying INFUSER to an instruction-finetuned anchor and augmenting it with rule-verifiable RLVR data, further demonstrate the flexibility and generalizability of the framework. Code is available at https://github.com/FFishy-git/INFUSER.

We study policy optimization for online episodic tabular Markov decision processes with unknown transition kernels, aiming for best-of-both-worlds guarantees together with data-dependent regret bounds. Recent work (Dann et al., 2023; Li et al., 2026) has shown that policy optimization can adapt to both adversarial and stochastic losses with first-order, second-order, and path-length bounds, but only under known transitions, leaving open whether such data-dependent guarantees are achievable by policy optimization when the transition kernel is unknown. We resolve this by developing a new algorithm based on optimistic follow-the-regularized-leader that attains these guarantees under unknown transitions. The key ingredient is a new design of optimistic $Q$-function estimators together with a data-dependent transition bonus that controls estimator bias through the loss-prediction error. Our analysis further identifies an unavoidable transition-dependent complexity term that captures the intrinsic cost of estimating the transition kernel. As a result, we obtain first-order, second-order, and path-length bounds with the transition-dependent complexity term while simultaneously achieving gap-dependent $\mathrm{polylog}(T)$ regret in the stochastic regime.

Abstractitations, recent work explores online RLHF that iterates between generating on-policy responses and collecting preferences [Lee et al., 2024, Park et al., 2022]. Among online The Self-Improving Alignment (SAIL) algorithmapproaches, SAIL reduces a bilevel alignment formulation addresses distribution shift by reducing a bilevelto a computationally efficient single-level surrogate and formulation of the problem to an efficient, single-reports strong empirical gains [Ding et al., 2024]. Empirically, SAIL has demonstratedisting online pipelines are largely heuristic and do not anastrong performance on this task. However, a for-lytically control the distributional shift induced by iterative mal analysis of its convergence properties has beendata collection [Chakraborty et al., 2024, Shen et al., 2024], lacking. We identify a key theoretical challenge: which has been linked to suboptimal performance in practice the standard SAIL objective function is not guar- [Sharma et al., 2024]. To address this limita-A growing line of work argues that the coupling between tion, we propose a regularized objective, SAILreward learning and policy updates is fundamentally bilevel and should be modeled as such [Chakraborty et al., 2024].RevKL, which incorporates a reverse KullbackAs a follow-up, Ding et al. [2024] reduces the bilevel align-Leibler (KL) divergence penalty to improve the optimization landscape. Our central theoretical con-ment objective to a tractable single-level surrogate and retribution is to prove that this regularized objectiveports strong empirical gains, yet it lacks formal convergence satisfies the Polyak-Lojasiewicz (PL) conditionguarantees. Related theoretical analyses in bilevel/RLHFstyle problems exist [e.g., Yang et al., 2025, Chakrabortywithin a bounded parameter space. We establish et al., 2024, Gaur et al., 2025], yet they either focus onglobal convergence guarantees, achieving a nearlinear sample complexity.

Accurate energy demand forecasting is essential for the reliable operation and planning of modern sustainable energy systems. Spatial-temporal graph neural networks (STGNNs) have recently achieved strong performance in point forecasting by jointly modeling temporal dynamics and relational dependencies across interconnected energy nodes. However, in real-world energy systems, accurate point forecasts alone are insufficient, as operators also require reliable uncertainty estimates to support risk-aware decision-making, grid stability, and operational planning under uncertainty. Conformal prediction provides a principled and model-agnostic framework for uncertainty quantification with statistical coverage guarantees, making it particularly attractive for safety-critical energy applications. However, existing conformal prediction approaches often fail to fully capture the complex spatial-temporal structure of energy systems. To address these limitations, we propose STOIC (Spatial-Temporal Graph Conformal Prediction with In-Context Learning), a novel framework that integrates graph-based forecasting with the zero-shot calibration capabilities of tabular foundation models. STOIC first generates point forecasts using an STGNN and subsequently reformulates spatial-temporal residuals into a tabular representation suitable for in-context learning. Leveraging a tabular foundation model, STOIC calibrates prediction intervals without task-specific retraining, effectively capturing both sequential and relational dependencies. We evaluate STOIC on five diverse benchmarks, including synthetic simulations as well as real-world electricity and district heating networks. Across all datasets, STOIC consistently outperforms existing conformal prediction baselines, delivering more reliable and robust uncertainty estimates for complex graph-structured energy time series.