AI Sec Watch

OpenAI paused its Stargate UK project, a planned multibillion-pound datacentre investment in Britain, citing regulatory concerns and high energy costs in April. An investigation revealed that OpenAI apparently never visited the key site in North Tyneside, and £20 billion of the £30 billion in investment that the UK government promoted appears to have been speculative rather than confirmed, raising questions about whether the project was primarily a publicity announcement rather than a genuine development plan.

Fanfiction communities are attempting to identify and remove works created using generative AI (large language models like Claude and ChatGPT that can create text automatically). However, the detection methods being used are unreliable and risk falsely accusing human writers of using AI.

The article explores whether AI language models (LLMs, which are trained systems that predict and generate text) can write fiction indistinguishable from human authors, amid growing concerns about AI use in publishing and media. It examines what linguistic features actually differentiate human writing from machine-generated text, drawing on perspectives from linguists and established novelists like Jennifer Egan and Jeanette Winterson.

Picklescan (a security tool that checks pickle files, which are Python files that serialize and deserialize objects) before version 0.0.33 fails to detect a specific dangerous code gadget called numpy.f2py.crackfortran.getlincoef that can hide in pickle __reduce__ methods (special functions that control how objects are reconstructed). This allows attackers to create malicious pickle files that execute arbitrary code when opened, potentially compromising shared model files in supply chains.

picklescan (a tool for detecting malicious code in pickle files, which are Python serialized objects) before version 0.0.30 has a vulnerability that allows attackers to hide malicious code in pickle files using a specific method (idlelib.run.Executive.runcode in reduce methods). When these files are loaded using pickle.load, the hidden code executes automatically, enabling RCE (remote code execution) and potential supply chain attacks on systems using PyTorch models.

This research proposes a framework for making AI hiring systems more fair by addressing algorithmic bias (when AI systems make systematically unfair decisions against certain groups). The study analyzes real applicant data and finds that language differences in interviews and how interview questions are structured can cause unfair outcomes, but these problems can be reduced by modifying linguistic features (the words and language patterns used) and making interview questions more consistent across all applicants.

A vulnerability in Keras (a machine learning library) version 3.14.0 allows attackers to run arbitrary code by exploiting how the Lambda layer deserializes data (converts stored data back into usable form). The bug occurs because the safety check treats an unset value the same as a deliberately disabled one, allowing malicious bytecode (low-level machine instructions) to execute when functions like `keras.layers.deserialize()` are called without proper safety protections.

Online predators are using nudification apps (AI tools that digitally remove clothing from images) to create fake sexual content of children from innocent photos posted online. The Report Remove service helps victims report and remove these explicit images from social media, but the widespread availability of these AI tools means children can be targeted without directly contacting criminals.

NSW government officials initially expressed enthusiasm about OpenAI opening a Sydney office, but removed language saying they were "absolutely thrilled" after staff members joked about dystopian AI scenarios similar to the Terminator films' Skynet (a fictional AI system that becomes hostile to humanity). The incident reflects some caution in how government communicates about AI expansion, even as it publicly encourages the technology.