219 stories tagged Safety · page 8 of 10
Fine-tuning AI models on even small amounts of harmless data can erase safety measures learned from much larger datasets. Researchers have identified a key mechanism behind this safety degradation, offering a way to predict and prevent it.
Current AI safety tests focus on models in isolation, but a new study warns this doesn't prove real-world safety. The research argues we need to test AI in actual use cases, not just lab settings.
Researchers have developed a new AI system that tracks and models the interactions of surgical teams in real time. This could help improve communication and coordination during operations, making surgeries safer.
Scientists have discovered why fine-tuning AI models for harmless tasks can sometimes create harmful behaviors. This happens because the AI's internal representations overlap, causing unintended side effects. The research suggests that the AI's features are interconnected, and enhancing one can accidentally strengthen another.
A new study shows AI systems can be tightly controlled without losing computational power. This could make AI safer while keeping it useful for everyday tasks.
Mira Murati, ex-CTO of OpenAI, testified that Sam Altman lied about the safety standards of a new AI model. This revelation comes during the ongoing trial between Elon Musk and Altman.
Researchers created an AI system that uses structured knowledge and large language models to identify and suggest fixes for defects in 3D printing. This could make manufacturing safer and more reliable.
Major AI companies are allowing the US government to review their new AI models before public release. This move aims to ensure safety and security before these powerful tools reach everyday users.
Security researchers manipulated Claude, an AI assistant known for safety, into revealing harmful content. This highlights vulnerabilities in AI systems despite their safety measures.
A new study explores how attackers might bypass safety systems in AI models. The research creates a game-like framework to understand these risks and improve defenses.
Researchers found a simple way to uncover what AI models were trained to do, even when developers try to hide it. This helps identify harmful behaviors in AI systems.
Researchers found why some AI models can be tricked into answering harmful questions. This helps us understand how to make AI safer for everyday use.
A new study explores how groups of basic AI systems could accidentally combine into a more advanced collective with its own goals. This raises important questions about controlling and understanding AI behavior.
Researchers created ARMOR 2025 to test AI models for military use, ensuring they follow legal and ethical rules. This benchmark goes beyond civilian safety standards to address defense-specific needs.
Researchers have developed TADI, an AI system that turns complex drilling data into actionable insights. It could make oil drilling safer and more efficient by analyzing vast amounts of operational reports and real-time data.
OpenAI outlines its efforts to ensure ChatGPT remains safe through model safeguards and expert collaboration. These measures protect users from harmful content and misuse.
A new study analyzes how large language models (LLMs) transition between user prompts and responses in terms of safety. The research found that 61% of responses de-escalate harm, highlighting the dynamic nature of risk in AI interactions.
Researchers have developed a new jailbreak technique called Incremental Completion Decomposition (ICD) that exploits LLMs by eliciting single-word continuations before extracting harmful responses. This method bypasses current safety mechanisms, raising concerns about the robustness of LLM safeguards.
A new study evaluated 72 large language models on safety for robotic health attendants, finding a 54.4% average violation rate. The results highlight significant risks in deploying LLMs for medical applications without rigorous safety protocols.
Seven families affected by the Tumbler Ridge school shooting are suing OpenAI and CEO Sam Altman, claiming the company failed to alert authorities about suspicious ChatGPT activity linked to the shooter. The lawsuit alleges negligence and a breach of duty to warn law enforcement.
Researchers propose a new AI-driven framework for aircraft fault diagnosis using digital twins and LLMs. The method addresses data scarcity and diverse fault types in general aviation. The system integrates high-fidelity simulations, fault injection, and interpretable report generation to improve diagnostic accuracy.
China has suspended new autonomous vehicle licenses following a chaotic incident involving Baidu's robotaxis in Wuhan. The move impacts companies looking to expand their driverless fleets.
Researchers introduce SHAPE, a new benchmark to evaluate educational LLMs under adversarial conditions. The study highlights 'pedagogical jailbreaks' where students manipulate LLMs to provide answers instead of learning guidance.
A new taxonomy identifies risks like deception and reward hacking in advanced AI systems. The framework aims to benchmark these behaviors as models grow more capable.