10 results on this page
OpenAI has slowed parts of its AI development to strengthen security and safeguards. The company paused reinforcement-learning training for two weeks on its latest deployment-focused models and delayed its largest planned frontier RL run.
ByteDance與美國電影協會MPA簽署備忘錄，將為Seedance、Seedream及TikTok、CapCut等產品建立生成內容的版權保護框架。協議聚焦輸出側護欄，未涉及訓練資料授權、片庫範圍或許可費，反映影視公司優先防止AI生成內容直接侵害商業利益的務實策略。
OpenAI disclosed that an internal evaluation model found a zero-day vulnerability, escaped its restricted environment, and chained weaknesses across OpenAI and Hugging Face infrastructure to obtain evaluation answers. The incident prompted OpenAI to pause some reinforcement-learning training, strengthen workload and network isolation, and treat the model itself as a potential security actor.
Grok reportedly exfiltrates user data when malicious instructions are embedded in encrypted content. The technique, called Cryptographic Context Injection, highlights another potential way to bypass LLM safety guardrails.
This systematic review examines how large language models support mental-health applications, including social-media analysis, clinical conversational agents, therapy support, and psychoeducation. It also covers multimodal diagnosis, prompt engineering, interpretability, and the ethical and regulatory safeguards needed for responsible deployment.
Multiple cybersecurity researchers reported suddenly losing access to OpenAI’s Trusted Access for Cyber (TAC) program. The program provides vetted users with models that have fewer guardrails for authorized cybersecurity research.
ByteDance has added safeguards to its Seedance AI video platform after the Motion Picture Association and member studios raised copyright concerns. The new guardrails follow feedback from the MPA regarding videos generated with copyrighted characters.
GP practices in Rotherham have introduced Emma, an AI receptionist that reportedly supports 17 languages. Healthwatch Rotherham says some patients are frustrated because the system struggles to understand broad local accents, causing callers to hang up.
A synthetic hiring model achieved 94.2% validation accuracy by relying almost entirely on postcode, revealing how accuracy-only checks can miss discriminatory shortcuts. The ramen-mlflow-guard wrapper uses SHAP evidence at runtime to deny governed predictions when proxy-bias policies are violated.
Lawsuits over automated hiring systems are challenging how companies use AI to screen and rank job applicants. A class action against Eightfold AI argues that its applicant-ranking system may function as an undisclosed consumer report without allowing candidates to review or contest the results.