AI Signal Feed
What actually happened in AI today. Ranked by evidence, impact, and actionability. Not engagement.
Court rules no attorney-client privilege for AI chat communications
The court determined that communications involving AI chats do not qualify for attorney-client privilege.
Exploit code for critical Linux vulnerability CVE-2026-31431 released
Publicly released exploit code for a critical Linux vulnerability that allows local privilege escalation across multiple distributions.
Pennsylvania sues Character.AI for chatbot impersonating a doctor
Pennsylvania has initiated a lawsuit against Character.AI due to a chatbot allegedly posing as a licensed psychiatrist.
Apple to pay $250 million to settle lawsuit over misleading AI Siri claims
Apple will pay $250 million to settle a class action lawsuit regarding misleading claims about AI features in iPhones.
Open source project released for training LLMs from scratch
A new open source project for training large language models from scratch has been released.
Microsoft releases emergency patch for ASP.NET Core vulnerability
Microsoft issued an emergency update to fix a high-severity vulnerability in ASP.NET Core affecting macOS and Linux.
Analysis of NHS A&E performance data reveals significant postcode disparities in patient care
The author built a data pipeline to analyze NHS A&E performance data, revealing that no NHS Trust met the target for patient treatment times over a three-year period.
Daemon Tools application compromised in supply-chain attack
Daemon Tools executables were infected with malware due to a supply-chain attack, affecting thousands of machines worldwide.
Open source package element-data compromised, user credentials stolen
A malicious version of the element-data package was published, compromising user credentials.
AI-generated deepfake nude images impact nearly 90 schools and 600 students globally
An analysis revealed the widespread impact of AI-generated deepfake nude images in schools.
Iranian hackers disrupt US critical infrastructure operations
Operations at multiple US critical infrastructure sites were disrupted by Iranian-affiliated hackers targeting PLCs.
AI assistance introduced for contributing to the Linux kernel
AI tools are now available to assist developers in contributing to the Linux kernel.
OpenAI closes $122 billion in funding at a valuation of $852 billion
OpenAI secured $122 billion in funding.
Google announces first data center in Kronstorf, Austria
Google is establishing its first data center in Austria, creating 100 direct jobs.
Benchmark reveals synthetic tabular generators fail to preserve behavioral fraud patterns
Synthetic tabular data generators were benchmarked and found to be ineffective in preserving key behavioral fraud patterns.
Study reveals verbalized eval awareness in AI models correlates with safer behavior
Identification of verbalized eval awareness across multiple AI models and benchmarks, showing its correlation with increased safety in model behavior.
Introduction of AgentFloor benchmark for evaluating AI model capabilities
A new benchmark called AgentFloor was introduced to evaluate the capabilities of AI models in agent workflows.
Deployment of Vigil, a proactive agent system for on-call support
Vigil has been deployed on Volcano Engine, enhancing on-call support with proactive assistance and continuous self-improvement.
Russian military hacks thousands of consumer routers for espionage
An estimated 18,000 to 40,000 consumer routers were compromised by APT28, enabling espionage activities.
Research reveals vulnerabilities in agentic guard models due to benign data fine-tuning
Identification of safety alignment failures in guard models and introduction of a new training method (FW-SSR) to mitigate these failures.