AI Sec Watch

OpenAI's chief revenue officer sent an internal memo to employees emphasizing the need to build a 'moat' (competitive advantages that make it hard for customers to switch to competitors) around its AI products and focus on enterprise clients, as users currently find it easy to switch between different AI models depending on which one performs best at any given time.

Gemini Robotics-ER 1.6 is an upgraded AI model designed to help robots understand and reason about the physical world, enabling them to complete real-world tasks with better spatial awareness and precision. The model improves on previous versions by enhancing capabilities like pointing (identifying and locating objects), counting, reading instruments (such as gauges), and detecting when tasks are complete. It is now available to developers through the Gemini API (an interface for accessing the model) and Google AI Studio.

Microsoft is testing ways to integrate OpenClaw-style features into Copilot, its AI assistant, to make Microsoft 365 Copilot run autonomously (without human intervention) around the clock and complete tasks for users. OpenClaw is an open-source platform that allows users to create AI-powered agents (software programs that act independently to complete goals) that run locally on a user's device. Microsoft's corporate vice president confirmed the company is exploring these technologies for enterprise use.

OpenAI's new revenue chief sent an internal memo highlighting a partnership with Amazon (a cloud computing company competing with Microsoft) as crucial for reaching enterprise customers, while acknowledging that its existing deal with Microsoft has constrained its ability to serve clients who prefer Amazon's AI platform called Bedrock (a service that provides access to major AI models). The memo reflects OpenAI's struggle to compete with rival Anthropic's Claude model in the enterprise market, where companies are investing heavily in AI.

A vulnerability in keras version 3.13.0 allows attackers to run their own code when a model is loaded, even when `safe_mode=True` (a setting meant to prevent unsafe operations). The problem occurs because the `TFSMLayer` class loads external TensorFlow SavedModels (pre-trained model files) without checking if they're safe, and doesn't properly validate file paths or configuration data.

Referring video object segmentation (RVOS, the task of identifying and outlining objects in videos based on text descriptions) is used in safety-critical applications like autonomous driving, but the deep neural networks that power these systems are vulnerable to adversarial perturbations (tiny, intentional changes to input data designed to fool AI models). This research demonstrates for the first time that RVOS models can be reliably attacked using a method called xM-ICM, which corrupts both visual and text information to mislead the models, and shows this attack works even when attackers have limited information about the system.

HKT-SmartAudit is a framework that creates smaller, faster AI models specifically trained to find bugs in smart contracts (self-executing code on blockchain networks). The framework uses knowledge distillation (a technique where a large, accurate AI model teaches a smaller model by sharing what it has learned), allowing these lightweight models to detect vulnerabilities effectively while using far less computing power than larger models.

FALCON-Net is a detection system designed to identify AI-generated images by analyzing their technical flaws. The system works by examining two key weaknesses in generated images: the lack of device-specific sensor noise (natural imperfections that real cameras add) and unnatural pixel intensity variations that result from oversimplified generation processes. FALCON-Net combines two analysis modules (one for noise patterns and one for local pixel variations) to reliably distinguish AI-generated images from real ones, even when tested on image generation models it wasn't trained on.

Federated learning (FL, where multiple devices train AI models together without sharing raw data) faces privacy risks because adversaries can extract sensitive information from model updates. FedNSA is a new protocol that combines differential privacy (adding mathematical noise to hide individual data patterns), encryption, and multi-party computation (MPC, a technique where multiple parties jointly compute results without revealing their individual inputs) to protect model updates while reducing the communication and computational burden that makes secure aggregation impractical on resource-constrained devices like smartphones.