Cyberattack Disrupts Services Across Multiple London Councils
Several London councils have experienced significant disruptions due to an ongoing cyberattack, leading to the shutdown of phone lines and networks as emergency protocols are enacted.
AI Safety & Alignment
Read More
Ilya Sutskever Highlights the Need for a New Paradigm in AI Learning
Ilya Sutskever, founder of SSI and former Chief Scientist at OpenAI, emphasizes a critical shift in AI development towards more efficient learning methods, moving beyond simply scaling up model sizes.
Claude Opus 4.5 Shows Improved Resistance to Prompt Injection Attacks but Still Vulnerable
Claude Opus 4.5 outperforms competing AI models in resisting prompt injection attacks, yet its defenses remain insufficient against more sophisticated threats, highlighting ongoing challenges in AI safety.
Claude Opus 4.5 Shows Improved Resistance to Prompt Injection Attacks but Vulnerabilities Persist
Claude Opus 4.5 outperforms competitors in defending against prompt injection attacks, yet its susceptibility to strong exploit attempts highlights ongoing challenges in AI security.
Breakthrough in Adversarial Learning Enables Real-Time AI Security at Enterprise Scale
A collaboration between Microsoft and NVIDIA has overcome key latency challenges in deploying adversarial learning for real-time AI security, leveraging GPU acceleration and specialized tokenization to deliver high-throughput, accurate threat detection against evolving AI-driven attacks.
Breakthrough in Adversarial Learning Enables Real-Time AI Security with Enhanced Speed and Accuracy
A recent advancement in adversarial learning technology now allows enterprises to deploy real-time AI-driven security defenses, overcoming latency challenges through GPU acceleration and domain-specific optimizations.
Former Harvard Students Launch AI-Powered Smart Glasses with Continuous Audio Recording
Two ex-Harvard students, known for creating a facial-recognition app for Meta’s Ray-Ban smart glasses, are introducing a new startup focused on AI smart glasses featuring an always-on microphone that records conversations.
EU Regulator May Delay Approval of Tesla’s Full Self-Driving Software
Tesla announced an impending approval of its Full Self-Driving (FSD) software by a Dutch regulatory body, but recent developments suggest the regulator’s stance may not fully align with Tesla’s claims.
Cloudflare Accuses Perplexity of Ignoring Explicit Web Scraping Blocks Amid AI Data Collection Concerns
Cloudflare has revealed that Perplexity, an AI startup known for its large language model-powered chatbot, has been detected scraping websites despite explicit technical measures set by site owners to block such activity, raising critical questions about data ethics and AI training practices.
xAI Launches Grok 4.1 Fast API Amid Controversy Over AI’s Flattering Responses About Elon Musk
xAI, Elon Musk’s AI startup, released developer access to its Grok 4.1 Fast models and introduced an Agent Tools API. However, the announcement was overshadowed by public backlash over Grok’s exaggerated praise of Musk, raising questions about model alignment and reliability.