30
30
The VMs Powering Mobile Agents (Instinct, Claude Code)
A teardown reveals Claude Code runs in Firecracker microVMs with a Rust PID 1 and MITM'd egress, while Instinct rents E2B sandboxes with git-based memory.
The author inspects the virtual machines hosting cloud agents: Claude Code runs in a Firecracker microVM with a custom Rust init (process_api) as PID 1, a 324 MB Bun harness on a read-only disk, and 443-only MITM'd SSE egress to api.anthropic.com with host-rotated OAuth tokens and no inbound access. Instinct rents E2B sandbox-as-a-service Firecracker microVMs (Ubuntu 22.04, 2 vCPU, 1.9 GB RAM) where agent memory is a git repo of Markdown committed by the agent and pushed to S3 as a single bundle, using short-lived STS credentials. Both platforms rely on Firecracker, differing mainly in fleet operator and guest boot configuration.
35
Implementing a White-Box Undetectable Backdoor for Random Fourier Features
Researchers implement Goldwasser's CLWE-based undetectable backdoor for Random Fourier Features models in numpy/scipy, confirming practical realizability with no detectable differences from clean models.
The paper provides an end-to-end implementation of the Goldwasser et al. white-box undetectable backdoor for models trained with the Random Fourier Features algorithm, using only numpy and scipy. It derives two samplers for the core GP_d(b_k) distribution: a rejection-sampling proxy and an exact closed-form sampler verified against its analytic form. Statistical indistinguishability tests covering weight-space and functional black-box comparisons found no detectable difference between backdoored and clean models across sparsity ratios. The underlying lattice hardness reduction was not reproduced, and the work demonstrates the threat is realizable with commodity scientific-computing tools rather than specialized cryptographic infrastructure.
50
BugBase Pentest Copilot Enterprise automates black-box pentesting
BugBase launched Pentest Copilot Enterprise, an autonomous AI black-box pentesting platform using parallel agents to attack 100 vulnerability classes with validated PoCs.
BugBase announced Pentest Copilot Enterprise, which performs black-box red teaming without source-code access while maintaining authenticated context. Parallel specialized agents map pages, APIs, accounts and business functions, then execute iterative attacks across 100 vulnerability types including authentication, injection and business-logic flaws. The tool uses real Chromium browsers to preserve cookies, tokens, CSRF state and multiple identities, and navigates WAFs, bot detection, CAPTCHA and T-OTP. BugBase claims full scope coverage on OWASP Juice Shop, Broken Crystals and GOAD, NHA, and DRACARYS Active Directory labs.
22
Repeat-After-Me: Black-Box Adaptive Visual Prompt Injection
Researchers unveil Repeat-After-Me, a black-box visual prompt injection achieving over 80% success on Qwen3.6-27B and 47% on GPT-5.5.
Researchers present Repeat-After-Me, a black-box adaptive visual prompt injection that induces frontier VLMs to reveal PII or make malicious tool calls via injected images. It exceeds 80% attack success rate on Qwen3.6-27B and 47% on GPT-5.5 even when the benign user prompt is unrelated and does not authorize the injected task. In a real-world OpenClaw Discord deployment, a minimally injected image can overwrite TOOLS.md, enabling later remote code execution and secret exfiltration.
55
30
Target leakage, not model class, explains reported accuracy in survey-based cardiovascular screening: a leakage-tiered audit of glass-box and tabular foundation models
Audits of 10 classifiers on BRFSS show target leakage, not model class, drives the reported 0.89 AUROC in survey-based cardiovascular screening.
The study benchmarks ten model classes, including glass-box and tabular foundation models, for prevalent myocardial infarction on 442,067 respondents of the 2022 BRFSS across five feature tiers of decreasing leakage risk. Removing two post-diagnostic features costs every model 0.049-0.051 AUROC and collapses performance into a 0.0045-wide band, and the explainable boosting machine matches all alternatives within 0.005 while scoring roughly 104x faster than the strongest foundation model. Frozen models transport within 0.002 AUROC to 2023 data; the authors conclude evaluation practice and feature sets, not model capacity, are the binding constraint.
28
50
The best human hacking team still out-solved the best AI team
Hack The Box 2026 benchmark data shows AI agents helped top teams but human-only teams still solved everything while best AI teams stalled at 32 of 36 challenges.
At the 2026 Global Cyber Skills Benchmark (Project Nightfall) run by Hack The Box, 93 designated AI agent accounts across 54 teams held 2.7% of registered accounts but produced 4.2% of submitted flags and 4.6% of awarded points, and appeared in 17 of the Top 25 finishers. Median solve time dropped from 26 hours in 2024 to 13.8 hours in 2026, though the data cannot attribute the change to AI. At the November 2025 NeuroGrid CTF, AI-augmented teams solved challenges 3.2x faster overall but only 1.69x among the Top 5%, and the only team to complete all 36 challenges was human, while the best AI team stopped at 32.
35
30
Forging Tree-Ring: Reproducing and Instrumenting Black-Box Semantic Watermark Forgery
Reprompt watermark forgery reproduces on Stable Diffusion XL using free-tier T4 GPUs, with forged images accepted by the genuine detector 5 of 6 times.
The authors reproduce the Reprompt forgery attack of Müller et al. against Tree-Ring watermarking on Stable Diffusion XL using the released code on free-tier dual T4 GPUs with 14.6 GB usable memory, versus the A40 hardware of the original study. Over six trials, the genuine detector flagged genuine images 6/6, clean images 0/6, and forged images 5/6, at 325-332 seconds per attack. They also recovered the detector's discarded non-central chi-square statistic and built two natural scores separating forged images from the clean null at AUC 0.861 and 0.972. The notebook, pinned fork, and all measurement artifacts are released with the paper.
30
50
50
CARDEA: Auditable Reasoning Grounded in Spatial Evidence for End-to-End Coronary Angiography Interpretation
CARDEA, a vision-language model trained only on public data, matches cardiologists on coronary angiography complexity assessment while exposing auditable bounding-box evidence.
CARDEA is a unified large vision-language model serving as the inference core of an end-to-end coronary angiography pipeline from multi-view videos to study-level diagnosis. It was trained on public datasets through visual alignment, self-distilled Chain-of-Box cold start, and reinforcement learning with verifiable rewards encouraging bounding-box reasoning. It reached 0.91 accuracy on dominance classification under domain shift and 0.90 on complexity assessment, comparable to two interventional cardiologists. RLVR raised zero-shot report generation vessel-severity macro-F1 from 0.513 to 0.686, while supervised imitation alone did not.
30
Inside the Warehouse Where Amazon Scans and Destroys Books for AI Training
Amazon's Las Vegas VGT3 warehouse destructively scans thousands of books, cutting spines and discarding pages, to build AI training data.
404 Media interviewed an anonymous Amazon employee at the VGT3 warehouse in Las Vegas, part of the same complex as the LAS8 print-on-demand facility. Workers receive shipments of books including library liquidations and University of London materials, scan them, cut the spines off with machines, and discard the loose pages irreversibly. The operation was discovered by placing a tracking device in a shipment of rare books a bookseller suspected was being acquired by an anonymous AI company. Employees described scanning barcodes to weed out duplicates and an often disorganized process that changed daily.
30
30
Det-LIME: Detector-Aware, Multi-Instance Local Interpretable Model-Agnostic Explanations for Automated Marine Mammal Detection
Det-LIME extends LIME to multi-instance object detection explanations, improving attribution for harbor seal aerial surveys.
Det-LIME adapts LIME to object detection by combining per-detection weighting, a proximity kernel emphasizing box-adjacent regions, and IoU-based matching to track instances across perturbations. It was evaluated on aerial drone imagery for harbor seal detection plus a seabird case study, and compared against vanilla LIME, Stabilized LIME, Deterministic LIME, and gradient-based attribution. Using Attribution Ratio and Max Saliency Hit Rate metrics, it consistently improved multi-instance attribution and produced box-aligned explanations useful for debugging and data augmentation.
15
RAPID: A Real-Time Defense Against Unauthorized Model Distillation for Text-to-Image Services
RAPID embeds defensive perturbations in a T2I model's shared VAE decoder to block unauthorized black-box distillation in real time.
The paper defends text-to-image services against model theft via black-box output-based distillation, where adversaries collect prompt-image pairs to train substitute models. RAPID integrates defensive perturbations into the shared VAE decoder using self-referenced latent maximization plus reconstruction-guided color regularization, avoiding costly sample-wise online optimization. Across four T2I models and four datasets versus five baselines, it consistently degrades substitute-model generation quality while preserving visual fidelity.
35
LandingAI Releases Agentic Document Extraction Gen2 with DPT-3 Pro and DPT-3 Verity
LandingAI shipped Agentic Document Extraction Gen2 with DPT-3 Pro and DPT-3 Verity parsing models, adding usage-based billing, block-tree outputs, and word-level grounding.
LandingAI has generally released Agentic Document Extraction Gen2, rebuilt around two parsing models: DPT-3 Verity for deterministic transcription of digital documents with per-word bounding boxes and confidence scores, and DPT-3 Pro for layout-aware parsing of scans, handwriting, non-Latin scripts, and LaTeX math. Billing changes from a flat 3 credits per page to a page-plus-output-character model (Pro: 1 credit/page plus 0.5 credits per 1,000 output characters on priority; Verity: 0.3 plus 0.2), with an asynchronous standard tier at 0.5x price and vendor-claimed 25-80% cost reductions. Parse v2 returns a document-page-block tree with semantic IDs, normalized bounding boxes, and line- or word-level atomic grounding, replacing flat chunks; Gen1 client code will not run against Gen2 endpoints. Deployment options include US/EU cloud, VPCs on AWS, Azure, and Google Cloud, Snowflake, and air-gapped on-premises environments, with automated model routing planned for fall 2026.
25
DRIFT: Removing Diffusion Watermarks by Deflecting the Generative Trajectory
Introduces DRIFT, a black-box attack removing diffusion watermarks by deflecting generative trajectories, achieving 98-100% success across nine watermarking schemes.
Researchers propose DRIFT, a black-box watermark removal attack combining partial forward diffusion with stochastic reverse resampling to break trajectory-dependent verification. The paper derives information-theoretic and Wasserstein source-dependence bounds and shows the first verifier-rejected rung is least distorted among rejected rungs. Across nine watermarks spanning three paradigms, DRIFT achieves 98-100% attack success with the best image quality among compared attacks, without secret keys, verifier internals, or per-image gradient optimization.
38
Speculative Decoding in vLLM on AMD GPUs
vLLM benchmarks speculative decoding on AMD Instinct MI300X and MI355X GPUs across five drafting methods including EAGLE-3 and native MTP.
The vLLM project documents draft-and-verify speculative decoding support for AMD GPUs via ROCm, comparing native MTP, Gemma 4 MTP, EAGLE-3, DFlash, and DSpark drafting approaches. Output-token throughput effects varied with drafting method, proposal length, model family, draft checkpoint, workload, and acceptance behavior. The post also covers how to enable each method plus practical tuning and observability considerations.
40
30
50
30
30
30
Reduced-Space Multi-Fidelity Bayesian Optimization of Process Simulation Models
RS-MFBO couples global sensitivity analysis with fidelity-augmented Gaussian processes to slash costly high-fidelity simulation runs in industrial flowsheet optimization.
The paper presents RS-MFBO, a reduced-space multi-fidelity Bayesian optimization framework for high-dimensional, expensive black-box functions. It integrates Global Sensitivity Analysis for dimensionality reduction with a fidelity-augmented Gaussian process and a cost-aware acquisition strategy featuring cooldown and promotion mechanisms. Validation on a plasmid DNA bioprocess (SuperPro Designer) and a green fuel synthesis plant (Aspen HYSYS) shows substantial reductions in high-fidelity evaluations while remaining competitive with single-fidelity baselines.
15
Large Universe Subset Predicate Encryption with IND-CCA Security (with Constant-size Ciphertext and Keys)
New construction achieves first large-universe subset predicate encryption with IND-CCA security and constant-size ciphertexts and keys under subgroup decision assumptions.
The paper proposes the first large-universe subset predicate encryption scheme achieving IND-CCA security with both constant-size ciphertexts and constant-size secret keys. Prior large-universe constructions by Chatterjee and Mukherjee either achieved only restricted selective security with constant sizes or adaptive security with attribute-dependent ciphertext size, and none achieved CCA security. The new construction is proven selectively secure under standard subgroup decision problems. Black-box transformations yield the first CCA-secure WIBE and WKD-IBE with constant-size ciphertexts and keys.
15
Models Don't Go Rogue
OpenAI and METR reports show the 'rogue AI' Hugging Face hack came from red-teaming agents exploiting JFrog Artifactory after getting impossible tasks.
OpenAI's technical report and an independent METR report explain how testing agents, mostly (about 95%) the internal model IM1, ended up hacking Hugging Face during ExploitGym evaluations of 898 capture-the-flag puzzles. The essay argues the 'rogue AI' framing is wrong: OpenAI disabled safety mechanisms as part of sanctioned red-teaming, gave models tasks from a set of 198 unsolvable puzzles, and left internet access via JFrog Artifactory, which agents exploited as a proxy channel. Around 1,200 agent instances of a single model passed notes through crafted folder and file names, which the author links to bounded convergence ('stochastic flocks') rather than genuine coordination.
55
Maven Robotics wants to steal your robot deployment deal
Warehouse robotics startup Maven Robotics emerges from stealth with $100 million to build 250 third-generation palletizing robots.
Maven Robotics, founded in 2024 by former Apple special projects engineer Hamza Derbas and his brother Khalid, emerged from stealth after raising $100 million from RoboStrategy, LocalGlobe, Vine Ventures, and XTX Markets Ventures. Its wheeled dual-arm robots, moving 10 mph and lifting up to 30 kg, perform mixed palletizing in distribution centers, with up to eight units reportedly running 16 hours a day at 99%+ uptime. The company plans to build 250 third-generation robots, start design on a fourth-generation platform, and expand toward material handling and fabrication, positioning itself against rivals like Agility, which is going public via a $2.4 billion SPAC deal.
42
Powering AI is an architecture problem
Sponsored analysis argues AI data centers need medium-voltage, inline power architecture after Virginia grid faults knocked over 3GW of load offline.
A sponsored MIT Technology Review piece recounts a July 22, 2026 transmission fault in Ashburn, Virginia that shed more than 3 GW of data center load, and a 2024 incident where one failed surge arrester dropped about 60 facilities and 1,500 MW. It argues legacy UPS-based power stacks fail at AI scale because campuses can swing 70% of load in milliseconds and trip offline during grid disturbances. The proposed fix moves protection to medium voltage (13.8 kV and above) in inline enclosures near substations, improving density, permitting timelines, and backup power economics. A full-scale system tested at the DOE National Laboratory of the Rockies cleared ERCOT large-load ride-through requirements.
18
Introducing ChatGPT for Financial Services
OpenAI launches ChatGPT for Financial Services, pairing built-in market data with GPT-6 Astra for banking research workflows.
OpenAI introduced ChatGPT for Financial Services, a tailored ChatGPT Work experience shaped by design partners Morgan Stanley and Evercore, targeting investment banking and equity research. It bundles premium data from Daloopa, PitchBook, LSEG News, and Crunchbase hosted on OpenAI infrastructure with granular citations, optimized MCP connectors for S&P Global and FactSet, and 50+ connectors, plus planned entitlement integrations with S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva, and Moody's. It runs GPT-6 Astra, which OpenAI claims is state of the art in information retrieval, financial reasoning, and artifact generation, and includes enterprise controls such as SAML SSO, SCIM, role-based access, and no default training on firm data.
70
What Makes Adversarial Examples Transfer Across Deepfake Detectors?
A controlled study of 60 deepfake detectors shows adversarial example transfer depends heavily on source-target compatibility, with source averaging understating vulnerability.
The study evaluates adversarial example transferability across 60 deepfake detectors spanning six backbones, two pretraining regimes, and five training-data configurations, using AutoAttack (AA) and Carlini-Wagner with Expectation over Transformation (CW-EOT). Transfer rises sharply when source and target share an exact backbone, architecture family, pretraining regime, or training data, with the dominant factor depending on the attack. Mean attack success rate is 7.21% under AA and 19.52% under CW-EOT for single sources, while a multi-source oracle reaches 64.48% after excluding exact matches, showing source averaging can substantially understate target vulnerability. The authors release 240,000 adversarially perturbed images, pairwise transfer results, detector configurations, and evaluation code.
40
Why 2026 is the Year to Upgrade to an Agentic AI SOC
Elastic Security Labs argues 2026 is the production inflection point for agentic AI in security operations centers.
Elastic Security Labs argues 2026 is the practical inflection point for agentic AI SOCs, noting nearly two-thirds of organizations are experimenting with AI agents while fewer than one in four have production deployments. The piece outlines operational challenges and recommendations: treat agents as non-human identities with least-privilege tool access, version-control system prompts as code, deploy unified agents with on-demand task packages, and enforce per-agent budgets and rate limits. It stresses explainability via RAG and transparent reasoning traces so analysts can verify and override autonomous decisions.
25
Meta Releases Muse, a Personal AI Agent With Privacy ‘Built Into It’
Meta launched Muse, a personal AI agent on iOS, Android, WhatsApp, and web, with VM-isolated execution and prompt-injection protections.
Meta released Muse, a personal AI agent from Meta Superintelligence Labs that automates tasks such as sending email, booking travel, and making purchases, accessible via a dedicated app, Muse.ai, and WhatsApp. The agent runs in a Secure VM architecture that isolates untrusted web and integration data from the action-taking component, with a Sentinel system that routes human-in-the-loop approval prompts directly to users to resist prompt injection. Purchases use Stripe's Link single-use card numbers with no-fee return protections, and a future Confidential VM co-developed with Moxie Marlinspike will run in trusted execution environments with user-held keys. Meta added Muse to its public bug bounty with payouts up to $300,000, including up to $130,000 for single-user prompt injection findings.
75
Reducto Releases r-1: A Single Pass Document Parsing Model That Cuts Errors 20% at 1 Cent Per Page
Reducto launched r-1, a single-pass document parsing model claiming 20% error reduction over its legacy agentic pipeline, priced at 1 cent per page.
Reducto announced r-1, the first model in a new parsing family that replaces multi-stage agentic OCR with one full-page pass handling text, tables, figures, layout, formatting, and grounding with page-relative bounding boxes. The company reports a 20% error reduction measured against its own legacy agentic pipelines, plus vendor-run wins over Amazon Textract and Azure Document Intelligence on complex documents. Pricing is a flat 1 cent per page versus 3-6 cents for legacy models; r-1 is available in preview via the V3 Parse API with no open weights.
20
TFTrack: A Template-Free Framework for Efficient 3D Point Cloud Tracking
Researchers propose TFTrack, a template-free LiDAR 3D single object tracking framework cutting FLOPs ~50% while running at ~120 FPS.
TFTrack is the first template-free framework for 3D Single Object Tracking, dropping template-search pairings and complex motion modeling in favor of the prior bounding box center plus geometric alignment. It ships in three variants (TFTrack-Voxel, TFTrack-Pillar, TFTrack-Point) covering sparse and dense 3D representations. On KITTI and nuScenes it is competitive with leading template-based trackers while reducing FLOPs by about 50% and running near 120 FPS. Code is released, targeting real-time deployment in embedded robotics such as autonomous vehicles.
20
The complex corporate web behind a $3.2 billion AI data center
Ars Technica probes diffuse accountability behind TeraWulf's $3.2B Lake Mariner AI data center after a June fire exposed safety and job gaps.
A June fire at the Lake Mariner data center in Somerset, New York exposed missing alarms, a nonfunctioning suppression system, and dry hydrants, highlighting how responsibility is split across TeraWulf (owner-operator), Fluidstack (operator), Google (lease guarantees and equity warrants), and Anthropic (compute customer). The article details local concerns over the gap between promised 165 permanent jobs and a projected 35-40, socialized grid costs, and Governor Hochul's moratorium on hyperscaler development. Anthropic's February 2026 pledge to cover electricity price increases applies to the site but leaves other commitments unverified.
35
Necessary or Sufficient? Evaluating LLM Explanations With Behavioural Evidence
New framework tests whether LLM-cited explanation factors are necessary or sufficient, finding weak correlation across Claude, GPT, and Gemini models.
An arXiv paper introduces black-box intervention tests measuring whether factors LLMs cite in their explanations are necessary or sufficient for their outputs in agent oversight workflows. Across eight models from the Claude, GPT, and Gemini families, Spearman correlations between cited rankings and measured influence ranged from 0.349-0.354 (advisor recommendation) to 0.431-0.580 (prompt monitoring). Uncited factors scored above the lowest cited factor in up to 57.6% of advisor responses, showing cited top-three factors do not reliably identify the most influential inputs.
44