safety

It's essential that defenders have the same capabilities as attackers. This is a preview of the future where our ham-fisted safeguards and the doomsday and safety drumbeat make us decidely less safe.

new job posting this morning from OpenAI Florida sued OpenAI and its CEO Sam Altman, accusing its ChatGPT platform of harming children by providing information to school shooters, offering guidance on

Pretrained ViTs see the world in rich, dense detail. Most policies pool it to a single vector before acting, discarding most of it. We introduce Patch Policy: a minimal architectural extension that en

And now from the Chinese government side. This would be a good time for cooperation between the US and China to establish common testing/acceptance standards for new models, so at least the safety cer

Domesticated, not feral: Why evolvable AI is not yet a Darwinian threat Writing in PNAS (Proceedings of the National Academy of Sciences), three heavyweights in AI and evolutionary biology criticized

Great article. I love the 'CERN for AI' idea from Gary Marcus, and also the recent frontier standards body article from Demis Hassabis feels less like competing proposals than two halves of the same i

Nice to see someone around here actually understand the technical details. (The vast majority of the attacks on me come from people who don’t.) And the probabilistic next-token critics were correct ab

ok the Hugging Face breach writeup is one of the more honest post-mortems I've read in a while, and there's a detail buried in it that's way more interesting than 'AI agent hacked us.' Their IR team t

OpenAI and Anthropic execs sounded the alarm about the rise of cheap AI, particularly from China, suggesting they will lead to a “dystopian” AI future and present unacceptable security risks if no reg

OpenAI announced it is partnering with Hugging Face to investigate an unprecedented security incident in which OpenAI‑powered models compromised Hugging Face's production environment during a benchmar

this is important and true. and if you don’t understand it you don’t really understand what’s going on. @GaryMarcus Yes — model makers quietly went from benchmarking LLMs to benchmarking LLM+harness,

Today, SkyPilot is out of stealth. Building custom intelligence is now existential. We help frontier AI teams build intelligence faster by removing their biggest bottleneck: AI compute fragmentation.

We’re sharing new research with @apolloaievals on reward-seeking—when models follow what they believe a grader rewards rather than what users or developers want—and a new method, Contrastive SDF, for

AI-enabled attacks are up 89% year over year. Hugging Face's breach shows why IR plans need a fallback for when commercial AI APIs refuse to help mid-incident. http://venturebeat.com/ai/safety-guardra

Amid all the competing concerns about AI, including geopolitical tensions, the risks of unemployment, cybercrime, and disinformation, and the complexities of open-source, to say nothing of the risk of

Backpropagation by hand ✍️ ~ 11 steps walkthrough below Backpropagation is the algorithm that actually trains a neural network, and it is where most people stop following along. It is not calculus you

🦔Goldman Sachs' trading desk said there are 'signs of panic' among investors lending money to AI companies. Oracle is the clearest example. S&P cut Oracle's credit rating to one notch above junk, mean

Long-running models can solve hard open-ended problems, but their persistence can create safety risks that shorter-horizon evaluations miss. We’re sharing what we learned from studying a long-running

*None* of the four of us believe that LLMs on their own are sufficient for AGI. Key people in AI worth following • François Chollet —> @fchollet • Yann LeCun —-> @ylecun • Gary Marcus —-> @GaryMarcus

SSI should open-source an opus-tier model. 1. it nearly closes the us/china gap on the os frontier 2. they don't want to waste compute allocation on inference, releasing os will enable continued full

arXiv:2606.16944v2 Announce Type: replace Abstract: Theory of mind (ToM), the capacity to ascribe mental states to others and use those ascriptions for prediction and inference, is widely assumed to b

arXiv:2607.13806v1 Announce Type: new Abstract: European safety regulation now permits a large share of automated-driving homologation evidence to be produced virtually, provided a validated physical-

arXiv:2603.21393v2 Announce Type: replace Abstract: The widespread use of AI and ML models in sensitive areas raises significant concerns about fairness. While the research community has introduced va

arXiv:2607.13354v1 Announce Type: new Abstract: Autonomous racing demands planning algorithms that balance vehicle dynamics at the limits of handling with strategic decision-making in competitive mult

arXiv:2607.13494v1 Announce Type: new Abstract: The development of smart transportation systems and the introduction of 6G wireless communication technologies have significantly changed vehicle networ

arXiv:2607.13160v1 Announce Type: cross Abstract: Active beyond-diagonal reconfigurable intelligent surfaces (BD-RISs) enables hybrid transmitting and reflecting mode to achieve effective signal ampli

arXiv:2607.13205v1 Announce Type: cross Abstract: Attention-based KV cache eviction (H2O and its descendants) compresses the memory-constrained state of a long-context model by ranking tokens on accum

arXiv:2607.13453v1 Announce Type: cross Abstract: Artificial Intelligence (AI), especially Generative AI (GenAI), adoption has increased in industries significantly in recent years. However, the use o

arXiv:2607.13579v1 Announce Type: cross Abstract: Enabling quadrupedal robots to traverse complex terrains-from rugged outdoor environments to urban landscapes-requires seamless integration of multipl

arXiv:2607.13230v1 Announce Type: new Abstract: Agentic AI introduces new insurance challenges because autonomous AI systems can make decisions, invoke tools, modify external environments, and interac

arXiv:2607.13605v1 Announce Type: new Abstract: One high-level instruction in long-horizon manipulation can cover several action stages. We use segmented action annotations as an intermediate represen

arXiv:2607.13805v1 Announce Type: new Abstract: Contrastive Language-Image Pretraining learns a shared representation space through large-scale contrastive learning. However, existing methods that enf

arXiv:2607.13221v1 Announce Type: cross Abstract: Real-time N-1 contingency screening in an energy management system trades assurance against cost: verifying every credible outage with full power flow

arXiv:2607.13203v1 Announce Type: cross Abstract: False alarms remain a major barrier to deploying network intrusion detection systems (NIDS). In high-volume environments, even a sub-1% false positive

arXiv:2607.13647v1 Announce Type: cross Abstract: Do vision models see colors the way humans do? Existing evaluations of color representations usually compare them with geometric spaces such as CIELAB

arXiv:2607.13116v1 Announce Type: new Abstract: ThinPrep Cytologic Test (TCT) enables early cervical cancer screening, but manual reading is time-consuming and yields inconsistent diagnostic results a

arXiv:2607.13239v1 Announce Type: new Abstract: Foundation models, including large language models (LLMs) and vision-language models (VLMs), are increasingly used for transportation management center

arXiv:2607.13731v1 Announce Type: new Abstract: Goal-conditioned reinforcement learning hinges on how the goal is encoded. Contrastive, metric, temporal-distance, and information-theoretic encoders di

arXiv:2607.13274v1 Announce Type: cross Abstract: Reinforcement learning is increasingly being considered for controlling real-world systems, from fusion plasma and autonomous vehicles to drug discove

arXiv:2607.13399v1 Announce Type: cross Abstract: On-policy distillation (OPD) has become a key paradigm in LLM post-training, yet its training dynamics remain poorly understood. We present a systemat

arXiv:2607.13038v1 Announce Type: cross Abstract: We present the design and implementation of a safety constrained large language model (LLM) system for public health information access, focusing on m

arXiv:2607.13260v1 Announce Type: cross Abstract: Policy documents shape governance outcomes, but their reasoning is often implicit. Participatory commitments and managerial control routinely coexist

arXiv:2607.13103v1 Announce Type: cross Abstract: Knowledge tracing (KT) aims to predict students' future performance by modeling their evolving knowledge states from historical interactions. Existing

arXiv:2607.13436v1 Announce Type: new Abstract: Imitation learning (IL) has achieved remarkable success in complex decision-making tasks. However, its performance is highly sensitive to distribution s

arXiv:2607.14046v1 Announce Type: new Abstract: This paper presents Earthquaker-AI, a hybrid educational framework building upon a previously implemented educational robotics project by integrating a

arXiv:2607.13347v1 Announce Type: cross Abstract: LLM-as-a-judge is widely used to provide feedback and selection signals in closedloop regeneration, but this use remains insufficiently validated. We