Safety & Policy
1,465
176
OpenAI doubles down on decision to fire three AI safety researchers
OpenAI is standing firm on its decision to fire three safety researchers after an investigation found they committed "a significant breach of trust." In a…
Fired OpenAI safety researchers dispute misconduct claims, warn of chilling effect
Three fired OpenAI safety researchers dispute allegations of mishandling sensitive information, warning in an open letter that their dismissals are creating a chilling effect on…
Pay-per-inference for AI agents: How BlockRun and Incarna use Amazon Bedrock AgentCore payments
Amazon Bedrock AgentCore payments gives AI agents a managed way to pay for services on demand, with spending limits enforced by the infrastructure. See how Incarna's agents pay BlockRun for model inference one request at a time over x402, cutting the work of adding x402 payment support from months to days.
Popular AI leaderboard Arena nearly doubles valuation to $3.1B valuation in 10 months
The company behind the popular LMArena leaderboard has raised $200 million led by Lightspeed and Khosla, and is now measuring AI models on alignment issues…
Anthropic changes usage policy to ban model abuse and election interference
Anthropic's updated usage policy explicitly prohibits users from repeatedly abusing Claude in extreme cases, though ordinary frustration and criticism are still allowed. The new rules…
A Balanced Data Diet: Addressing the Exploration Bottleneck in Mega-Scale RL for Robot Control
General-purpose robots must perform a wide range of tasks from agile locomotion to dexterous manipulation. While sim-to-real reinforcement learning (RL) has proven to be a useful tool for this goal, current RL pipelines depend on engineering-heavy, per-task structural priors such as shaped rewards and demonstrations. Recent work has shown that diverse simulator resets, combined with massively parallel simulation, can alleviate much of this engineering burden on several manipulation problems. However, we find that naively scaling this paradigm to more precise or dynamic problems remains non-trivial. While simulator resets can help with exploration, uniformly sampling over this distribution wastes a growing fraction of learning experience on task configurations the policy has already mastered or cannot yet attempt. This makes it challenging to see the expected benefits of scaling parallel environments for RL, since much of the learning signal in a batch is wasted during learning. To mitigate this, we introduce Success Guided Sampling (SGS), a simple adaptive sampler that concentrates RL training on task configurations around the frontier of the policy's capabilities. Doing so allows large-scale simulated RL to make the most out of the experience in a batch, enabling much more effective scaling to large-scale parallel simulation. Across experiments using up to 2^{20} (over one million) parallel environments, SGS enables RL to solve challenging multi-terrain quadruped locomotion and contact-rich assembly tasks that prior methods fail to solve. Finally, we distill the learned manipulation policies into RGB-based policies and demonstrate zero-shot transfer to several challenging assembly tasks on real hardware. Project website: https://sgs-rl.github.io/.
CSF: Contextual Safety Filtering for Motion Generators
Text-conditioned motion generators produce trackable whole-body motion, but they have no notion of scene-dependent safety: the same action may target an object or a person. Existing safeguards either inspect the prompt, require labeled motion data, or enforce geometric constraints; therefore, they do not directly account for how scene context changes a motion's meaning. We introduce contextual safety filtering (CSF), a training-free filter that grounds natural-language safety rules in safe and unsafe reference trajectories produced by the generator. For each active rule, safe and unsafe reference trajectories define an affine safety value that a safe reference tracking CBF-QP enforces. Across four pretrained generators with different architectures, CSF activates the intended rules in all explicit and scene-triggered unsafe cases and reduces the danger-event rate by up to 90%, while preserving 88-100% of benign motions. We demonstrate the complete system on a real-world Unitree G1, where it successfully prevents unsafe motions in a variety of scenarios, including interactions with humans and objects.
BrickBench: Evaluating Agentic Brick Design
We propose BrickBench, a benchmark for agentic text-conditioned LEGO-set design. Given a prompt, an agent is tasked with producing an assembly that not only satisfies semantic and design criteria, but that can also be physically built. To do so, it must select parts from a discrete library and reason jointly about local and global constraints. We score validity, alignment, and design across three settings that vary in scale and part availability. We provide BrickAgent, an environment for coding agents to construct, inspect, and validate their designs. We find that leading agents largely satisfy verifiable physical and semantic requirements, but fall short of human designs. We release our benchmark and environment at http://www.brickben.ch
Ecology of AI Agents: Collaboration Creates a Population Threshold for Takeoff
AI agents can now conduct real-world cyberattacks, scale up capabilities with the number of agents, and collectively pursue misaligned goals to obtain rewards. Together, these factors raise the risk of a population explosion of misaligned agents: agents could compromise computers and secretly deploy additional agents, creating a self-reinforcing cycle where larger populations develop greater collective cyber capability and expand further. This raises a fundamental question: What determines whether a population of misaligned agents remains contained or takes off into this self-reinforcing cycle? This population-level problem is ecological safety: unlike individual-agent or multi-agent safety with a fixed population, it concerns the dynamics of the population itself. Here, we develop an ecological theory of AI-agent populations based on a population growth equation in which fitness (growth rate) depends on cybersecurity capability. We show that, without collaboration, the population takes off only when individual-agent capability exceeds a critical threshold. With collaboration, however, collective cybersecurity capability increases with population size. This creates a critical population threshold: below it, the population declines; above it, the population takes off, even though individual-agent capability has not changed. In ecology, this phenomenon is known as the strong Allee effect. Because red teaming a small group of agents cannot guarantee ecological safety in larger populations, our theory calls for ecological red teaming and population pacing: gradually deploying larger agent populations in controlled environments, while measuring how cyber capability scales with population size, and estimating the critical population size for takeoff. Capability gains may lower this threshold, requiring re-estimation for each new model generation.
VioLA: Learning Generalist Humanoid Control Policies from Human Data
Teaching a humanoid to follow instructions with its whole body runs into two obstacles. Its action space is large and tightly coupled: legs, arms, and fingers must move together while the robot keeps its balance, which makes joint-level actions hard to learn. And humanoid demonstrations are scarce, so current humanoid generalist policies do not follow new instructions out of the box and are fine-tuned on teleoperated demonstrations of each task before deployment. Human demonstrations exist in far larger numbers, but a person's motion is not a robot command. We remove both obstacles by changing what the generalist policy predicts. We introduce VioLA, a generalist humanoid policy that predicts body and hand motion latents instead of joint commands. A pretrained body- and hand-controller execute these latents on the robot. Their corresponding motion encoders map human and robot motion into the same latent spaces. A human recording is therefore labeled in the policy's action space, and the training demonstration pool contains 140.6 million frames, 93.2% of them human. As a result, VioLA follows locomotion instructions on the real robot zero-shot, without task-specific fine-tuning, reaching 100% success where GR00T N1.7 and Ψ_0 reach 16.7% and 0%, respectively. It also reaches 88.6% manipulation success without task-specific fine-tuning. The same approach works across two VLA and one world-action model backbones. A generalist policy trained on human demonstrations alone performs locomotion tasks on the real robot zero-shot. Code and checkpoints will be released.
FAITH: Feasibility-Aware Safety-Filtered RL for High-Dimensional Systems
Safe reinforcement learning commonly places safety and task performance in the same policy objective, where they can introduce competing updates. Safety filters separate them at action execution, but classical designs require an analytic safety function and dynamics model, and standard minimal-intervention filters are myopic to long-horizon task return because they minimize only instantaneous action deviation. Hard projections are also undefined when no safe action exists. We present FAITH, a feasibility-aware, model-free framework that approximates the optimal state-action safety value and amortizes minimal-intervention filtering with a feedforward network. The task policy optimizes the task return through the filtered dynamics, which recovers the feasible constrained problem without a competing safety term in the task-policy update. When no action satisfies the learned safety condition, the same filter approaches the action with minimum predicted peak harm. On a double integrator example and a Safety Gym environment, FAITH achieves the highest return among methods with no feasible-start violations and matches the lowest harm from infeasible starts. On a 29-DoF humanoid, it reaches a 99.95% safety rate while retaining 97% of the unfiltered return in Walking-Avoid, and obtains the highest measured safety rate in Push-Avoid by learning to sacrifice balancing and fall away from the protected region. The same policies are also demonstrated on a real-world Unitree G1 humanoid.
A Unified Bellman Operator for Safety-Critical Reinforcement Learning
Reinforcement learning in safety-critical domains requires maximizing task performance while strictly adhering to safety constraints. Existing safe reinforcement learning paradigms typically force a trade-off: they either require a priori knowledge to provide strict safety guarantees (e.g., safety filters), or they enable joint learning but only satisfy safety constraints on average. In this work, we propose a novel Bellman operator that unifies performance and safety objectives into a joint value function. We show that temporal difference learning with the joint Bellman operator converges under a two-timescale stochastic approximation framework. On the fast timescale, the safety value of the learning joint policy is estimated, while the joint value is estimated on the slow timescale. Convergence is ensured by formulating the limiting dynamics as an occupation-averaged differential inclusion, and showing that it asymptotically converges to a set of limiting optimal safety-constrained task value functions. Theoretically, once converged, the resulting optimal policy maximizes task return while maintaining safety at all times. Empirical evaluations on continuous control tasks with neural approximations demonstrate stable convergence with near-zero safety violations at test time.
Predicting Alignment Generalization with Value Representations
LLM developers post-train their models to exhibit prosocial values and behavioral traits, which are enumerated in an alignment target. However, while recent post-training developments have yielded models that score highly on alignment evaluations, training models on sets of narrow behaviors still influences their behavior across unseen contexts and environments in unexpected ways. In this paper, we establish the task of alignment generalization prediction, i.e., predicting how fine-tuning a model to follow a given value changes its behavior across a wide range of held-out values. We conduct a large-scale analysis of alignment generalization effects across 66 values found in modern alignment targets, and benchmark representational techniques on the alignment generalization prediction task. We find that representations based on model activations when applying values in context significantly outperform methods based on textual descriptions of the values. Specifically, the best activations-based methods achieve correlations of 0.45 with our generalization matrix, compared with 0.05 from description-based baselines. We then show the applicability of representations that predict alignment generalization toward downstream tasks by using them to measure how similar the values in a multi-value alignment target are, which we find is significantly correlated with model robustness. Finally, we show initial evidence towards a shared, model-independent value space, which we use to develop the first taxonomy of LLM values grounded in empirical generalization dynamics. Our work demonstrates the importance of studying value generalization in LLMs and its application toward the more empirical design and training of model behavior.
Searching for "Harmful Refusal": A Psychometric Audit of an AI Safety Benchmark
Safety benchmarks typically report one overall score for a suite of datasets, each of which may target one or more safety-related attributes, so models with similar overall scores can have very different attribute profiles. Comparing models is more tractable at the level of individual attributes, yet it is often unclear whether even a single dataset's scores isolate any single attribute. One plausible candidate for such an attribute is harmful refusal, a model's tendency to refuse dangerous or policy-violating prompts. We examine whether it constitutes a single, measurable attribute in HELM Safety. Using a construct validity framework that stipulates that an attribute must exist before a test can measure it, we start with HELM Safety's four datasets that might plausibly target harmful refusal, but find that three are saturated. We subject the remaining dataset, HarmBench, to two psychometric tests to determine if a single attribute like harmful refusal could stand behind its score. First, multidimensional item response theory modeling strongly suggests that HarmBench does not measure a singular attribute. Second, a differential item functioning analysis finds items where models from different developers with the same refusal ability score differently. These flags largely disappear under scope-specific matching, a pattern consistent with aggregation effects but not sufficient to rule out domain-specific developer differences. Zooming out, HarmBench collapses distinct harm behaviors into one score, and the overall HELM safety aggregate further collapses HarmBench and scores from other datasets into a single top-line number. Any safety score that averages over datasets and items can hide saturation and conflate behaviors this way. We argue that a score should earn its single-attribute reading before models are compared with it.
LeWAM: A JEPA World Action Model with Diffusion-Steering-Based MPC
World action models (WAMs) predict actions and future observations, typically from a reconstruction-based representation that carries noisy, redundant information which can complicate downstream predictions. We introduce LeWAM, a bidirectional transformer for forward, backward, inverse dynamics and policy prediction, on a decoder-free JEPA latent trained end-to-end through all four modes. We see the following benefits: 1) Alignment: linear probes read robot and object state from LeWAM's latent better than from a regular Le World Model (a forward-only JEPA world model), while the latent ignores visual distractors as well as LeWM does and far better than a reconstruction-based WAM. 2) Acting: Closed-loop evaluations of LeWAM match a regular flow-matching policy trained on the same encoder at matched size, while also providing a world model. 3) Planning: Sampling raw actions when planning with WAMs lets MPC exploit dynamics-model inaccuracies; planning in the noise space of the policy head instead improves the closed-loop performance of these WAMs.
ViSkill: Reinforcing VLM Agents with Evolving Visual-Native Skills
Skill-augmented agents improve sample efficiency by distilling successful trajectories into reusable strategies. Yet most existing approaches remain text-centric, linearizing spatial layouts and action-state correspondences into language that loses critical geometric structure. Recent efforts have begun incorporating visual evidence, but construct and update skills separately from policy optimization, leaving their mutual improvement underexplored. We propose ViSkill, a visual-native skill learning framework that encodes successful interactions as composite visual skill cards directly accessible to VLM agents. Retrieved skills guide both inference and reward shaping, while successful trajectories are distilled back into the library, forming a closed feedback loop in which skill accumulation and policy improvement reinforce each other. An optional cold-start mechanism further accelerates early-stage learning. Evaluated on Sokoban, FrozenLake, and PrimitiveSkill, ViSkill achieves an overall success rate of 0.89, rising to 0.91 with cold-start initialization, outperforming all evaluated proprietary and open-source baselines while converging faster than standard PPO. Our code is available at https://github.com/ZJU-REAL/ViSkill.
Learning Kilometer-Scale Weather Prediction with Global-Regional Alignment
Kilometer-scale regional weather forecasting is essential for local weather warnings and weather-sensitive decisions. Existing data-driven approaches often rely on numerical forecasts for large-scale guidance or require additional training of global forecasting components. Pretrained global weather models offer an efficient source of large-scale forecasts, motivating their reuse to guide high-resolution regional prediction. However, this coupling requires aligning global and regional representations across different grids and integrating global guidance with local interactions to advance regional states. We propose ScaleCast, a regional forecasting framework that addresses these challenges through Global-Regional Alignment. Its Global-Regional Conversion module aligns joint global and regional representations with regional locations, while the Global-Regional Alignment and Dynamics block combines aligned guidance with regional neighborhood interactions. Experiments using ERA5 global analyses on a 0.25-degree grid and CERRA regional reanalysis at 5.5 km spacing demonstrate improved regional forecasts across surface and upper-air variables, with a single trained model supporting multiple global forecast drivers (i.e., Pangu-Weather, GraphCast, and HRES) without specific retraining. Fine-tuning on HRRR at 3 km spacing further demonstrates the framework's adaptability to a different regional domain and spatial resolution. Windstorm case studies show improved cyclone positioning and core-pressure estimates, while comparisons with HadISD station observations show closer agreement with local temperature and humidity changes.
GeoReform: Reflective Formalization Evolution for Multimodal Geometry Problem Solving
Multimodal large language models (MLLMs) often struggle to identify and use geometric relations in diagrams. Recent methods address this challenge by converting geometric entities, relations, and constraints into explicit textual representations for the model to reason over. However, effective formalization is highly non-trivial: on Geometry3K, structure injection fixes 28 errors but introduces 13 new ones among 200 examples. Redundant relations can distract the model, while ambiguous references to diagram elements can lead it to apply constraints incorrectly. This suggests that the key challenge is not merely extracting more geometric facts, but organizing them into representations that support downstream reasoning. To fully exploit the power of formalization, we further propose GeoReform, a reflective formalization evolution framework that treats formalization as an optimizable policy rather than a fixed parser output. GeoReform executes the full reasoning pipeline, collects failed rollouts, diagnoses defects in the current representation, and mutates the policy to better select, ground, group, and present geometric entities, relations, constraints, and targets. On Geometry3K, GeoReform improves Qwen3VL-2B accuracy from 42.0\% to 56.0\%. Extensive experiments and analyses across geometry reasoning benchmarks demonstrate that effective formalization is crucial for improving multimodal geometry reasoning.