In enterprise settings, organizational data is segregated, siloed and carefully protected by elaborate access control frameworks. These access control structures can completely break down if an LLM fine-tuned on the siloed data serves requests, for downstream tasks, from individuals with disparate access privileges. We propose Permissioned LLMs (PermLLM), a new class of LLMs that superimpose the organizational data access control structures on query responses they generate. We formalize abstractions underpinning the means to determine whether access control enforcement happens correctly over LLM query responses. Our formalism introduces the notion of a relevant response that can be used to prove whether a PermLLM mechanism has been implemented correctly. We also introduce a novel metric, called access advantage, to empirically evaluate the efficacy of a PermLLM mechanism. We introduce three novel PermLLM mechanisms that build on Parameter Efficient Fine-Tuning to achieve the desired access control. We furthermore present two instantiations of access advantage-(i) Domain Distinguishability Index (DDI) based on Membership Inference Attacks, and (ii) Utility Gap Index (UGI) based on LLM utility evaluation. We demonstrate the efficacy of our PermLLM mechanisms through extensive experiments on five public datasets (GPQA, RCV1, SimpleQA, WMDP, and PubMedQA), in addition to evaluating the validity of DDI and UGI metrics themselves for quantifying access control in LLMs.

The text produced by language models (LMs) can exhibit specific'behaviors,' such as a failure to follow alignment training, that we hope to detect and react to during deployment. Identifying these behaviors can often only be done post facto, i.e., after the entire text of the output has been generated. We provide evidence that there are times when we can predict how an LM will behave early in computation, before even a single token is generated. We show that probes trained on the internal representation of input tokens alone can predict a wide range of eventual behaviors over the entire output sequence. Using methods from conformal prediction, we provide provable bounds on the estimation error of our probes, creating precise early warning systems for these behaviors.

We introduce PurpCode, the first post-training recipe for training safe code reasoning models towards generating secure code and defending against malicious cyberactivities. PurpCode trains a reasoning model in two stages: (i) Rule Learning, which explicitly teaches the model to reference cybersafety rules to generate vulnerabilityfree code and to avoid facilitating malicious cyberactivities; and (ii) Reinforcement Learning, which optimizes model safety and preserves model utility through diverse, multi-objective reward mechanisms. To empower the training pipelines with comprehensive cybersafety data, we conduct internal red-teaming to synthesize comprehensive and high-coverage prompts based on real-world tasks for inducing unsafe cyberactivities in the model. Based on PurpCode, we develop a reasoning-based coding model, namely PurpCode-32B, which demonstrates state-of-the-art cybersafety, outperforming various frontier models. Moreover, our alignment method decreases the model overrefusal rates in both general and cybersafety-specific scenarios, while preserving model utility in both code generation and common security knowledge.

There's a mad dash to automate the world's most hated calls. You'll hear from an AI debt collector sometime soon. She introduced herself as Eve, but Ben knew right away that the voice on the other end of the line was a bot. She also knew how much money he'd owed a former landlord ($266). She didn't seem to know that he'd settled with a collection agency five months prior. Eve said she was an AI agent from ProCollect and was calling to collect a debt.

When you purchase through links in our articles, we may earn a small commission. At $110, the Tapo MagCam 4K C465 is at its all-time low at Amazon. This solar-powered Wi-Fi security camera is as convenient as they come. The Tapo MagCam 4K C465 security camera is high-def, easy to install, and now available for 21% off. That means you can score it for just $110 at Amazon right now, which matches the all-time lowest it's ever been.

A recent attack on Sam Altman's home and OpenAI offices has put corporate security under renewed scrutiny. Records reveal how much some tech firms spend to arm up. Elon Musk called violent crime in San Francisco " horrific " and moved the offices of his social media business X outside the city in 2024 because of safety and business considerations. Other local tech companies have attempted to address their security concerns by partnering directly with cops. Airbnb and Salesforce are among businesses that for years have contracted San Francisco police to protect their offices on a regular basis, according to public records obtained by WIRED.

We introduce the Hiremath Early Detection (HED) Score, a principled, measure-theoretic evaluation criterion for quantifying the time-value of information in systems operating over non-stationary stochastic processes subject to abrupt regime transitions. Existing evaluation paradigms, chiefly the ROC/AUC framework and its downstream variants, are temporally agnostic: they assign identical credit to a detection at t + 1 and a detection at t + tau for arbitrarily large tau. This indifference to latency is a fundamental inadequacy in time-critical domains including cyber-physical security, algorithmic surveillance, and epidemiological monitoring. The HED Score resolves this by integrating a baseline-neutral, exponentially decaying kernel over the posterior probability stream of a target regime, beginning precisely at the onset of the regime shift. The resulting scalar simultaneously encodes detection acuity, temporal lead, and pre-transition calibration quality. We prove that the HED Score satisfies three axiomatic requirements: (A1) Temporal Monotonicity, (A2) Invariance to Pre-Attack Bias, and (A3) Sensitivity Decomposability. We further demonstrate that the HED Score admits a natural parametric family indexed by the Hiremath Decay Constant (lambda_H), whose domain-specific calibration constitutes the Hiremath Standard Table. As an empirical vehicle, we present PARD-SSM (Probabilistic Anomaly and Regime Detection via Switching State-Space Models), which couples fractional Stochastic Differential Equations (fSDEs) with a Switching Linear Dynamical System (S-LDS) inference backend. On the NSL-KDD benchmark, PARD-SSM achieves a HED Score of 0.0643, representing a 388.8 percent improvement over a Random Forest baseline (0.0132), with statistical significance confirmed via block-bootstrap resampling (p < 0.001). We propose the HED Score as the successor evaluation standard to ROC/AUC.

You don't have to upload your video to the cloud or pay a monthly fee to secure your home. In the age of state surveillance, with big tech trampling our data privacy rights and gouging us for every penny, there are plenty of reasons to keep your security camera footage local. Whether you want to save money or ensure your video doesn't end up in the hands of persons (or AI) unknown, subscription-free security cameras are the way to go. The good news is that locally recording security cameras are better than ever. I've been testing security cameras for a decade, and the gap between the best cloud-connected and local cameras is closing. You don't necessarily have to give up the best features to shirk that subscription anymore.

Neural Information Processing Systems http://nips.cc/

When you purchase through links in our articles, we may earn a small commission. BotsLab 4-Cam W510 System review: This security package doesn't deliver Four 4K cameras, a base station with expandable local storage, and no subscription required, So, what's the catch? This four-camera system impresses with solid video quality and expandable local storage, but only when those cameras are in such close range that they probably won't provide full coverage of your property. Outfitting your home with outdoor security cameras can get complicated--and expensive--quickly. Anyone looking for a shortcut on both fronts might consider one of BotsLab's W510 kits, bundles consisting of up to six 4K outdoor pan/tilt security cameras, solar panels to keep each camera's battery topped off, and a base station with 32GB of onboard storage (expandable up to 16TB with a user-supplied 2.5 hard drive).