Research
Academic papers, new techniques, benchmarks, and theoretical findings in AI/LLM security.
Academic papers, new techniques, benchmarks, and theoretical findings in AI/LLM security.
2,033 items
Google DeepMind and Google Research, working with the National Hurricane Center, CIRA, the UK Met Office and other weather agencies, report that their WeatherNext AI model achieves state-of-the-art accuracy in predicting a cyclone's track, intensity and wind structure, giving forecasters on average an extra day of predictive accuracy. The team is open sourcing the WeatherNext 2 and WeatherNext Cyclones models, which predict 1,000 possible scenarios per cyclone.
The source introduces a Bayesian intent lattice for detecting metamorphic malware, a category that rewrites its own code to evade pattern-based detection. The article was published online on 31 July 2026 in Computers & Security by Muhammad Abdur Raafay.
This paper introduces CKICAS, the Community Kinetic Intelligent Complex Adaptive System, which extends adaptive systems principles to community contexts with informal governance. Developed through Action Design Research during COVID-19, it reported compressed decision-response cycles across five crisis contexts, including coordination across five jurisdictions. The authors state that systematic experimental evaluation remains future work.
A comprehensive survey of compression algorithms for language models appeared in ACM Computing Surveys, Volume 58, Issue 14, pages 1-35, in October 2026. The source text provides only the bibliographic citation and no abstract, method, or findings.
This is the release candidate (RC) 2026-07-28 revision of the Model Context Protocol, with the specification available in draft form and a changelog linked for details. The source states the specification is not final and that changes may still be introduced before the final release.
The Model Context Protocol has published a stable release of its 2026-07-28 revision. The specification is available on the official Model Context Protocol website, and a changelog describes the changes in detail.
This comparative case study examines how shifting teaching and learning logics and digital technologies co-transform higher education. The authors studied two public universities, one in Germany and one in Australia, and identified five conceptual themes of digital transformation, from a platformization approach to partnership models. These themes connect to three logic shifts: from mass production to mass personalization, from provider-centric to student-centric, and from students as consumers to students as prosumers.
This paper presents DeepForgeSeal, a deep learning framework for proactive deepfake detection using semi-fragile watermarks. A learnable watermark embedder operates in latent space, and Adversarial Reinforcement Learning (ARL) trains it against a curriculum of benign and malicious image manipulations generated by an adversarial attacker agent. On the CelebA and CelebA-HQ benchmarks, the authors report improvements of over 4.5% and more than 5.3% respectively under challenging manipulation scenarios.
OpenAI reports ten new results on open mathematics and theoretical computer science problems, each open for at least a decade. The results were achieved by an internal version of its next major model, Astra, at a token cost of roughly $2,000 at Sol API rates. Humans then prepared the arguments as manuscripts, and the model formalized each argument in a Lean certificate.
OpenAI analyzed more than 800,000 messages from U.S. ChatGPT users and found that 16.8% of work-related messages and 43.5% of occupation-specific messages concern tasks associated with another occupation, a pattern it calls task crossover. The report, Work at the Frontier: How AI is Expanding What People Do at Work, is the first in a new series tracking how AI changes work. Customer experience workers (77%), designers (75%), human resources workers (69%), legal workers (56%) and marketers (53%) show the highest shares of outside-occupation tasks.
The source is a journal citation for an ACM Transactions on Privacy and Security paper titled "Watermarking for Model Ownership Verification: Invisible at Deployment, Activated by Updates." It appears in Volume 29, Issue 3, pages 1-28, August 2026. The text provided contains no abstract, method, or findings.