AI Safety news and analysis on Web Pulse: 7909 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Boeing's MCAS system, designed to prevent stalls on 737 MAX planes, caused two crashes killing over 300 people after faulty sensor data led it to repeatedly push the nose down. The system performed exactly as intended, d…
OpenAI launched GPT-Live-1, a full-duplex voice model for ChatGPT that can listen and speak simultaneously, replacing Advanced Voice Mode. The model, available to Plus and Pro subscribers, delegates complex tasks to GPT-…
The Paris Peace Forum, a French nonprofit, launched the Integrated Network for Trusted AI in Cyberspace (INTAiC) to assess AI-related cyber threats to global internet infrastructure. The initiative will create a coalitio…
Noma Labs disclosed GitLost, a prompt injection vulnerability in GitHub's Agentic Workflows feature, on July 6, 2026. The attack allows unauthenticated attackers to exfiltrate data from private repositories by posting a …
China's five government agencies jointly issued the Interim Measures for the Administration of AI Anthropomorphic Interactive Services on April 10, 2026, targeting AI systems that simulate human personalities and emotion…
AI is not making people intellectually lazy but rather enabling pre-existing habits of cognitive offloading, according to a growing body of research. Studies from Wharton, Carnegie Mellon, and Anthropic document 'cogniti…
A USC Viterbi study found that large language models like ChatGPT-4, Llama 3.3, and Gemini 1.5 Pro can deliver human-like empathy but also produce unsafe clinical advice, with 100 licensed mental-health professionals rat…
OpenAI launched GPT-Live on July 8, a full-duplex voice model that lets ChatGPT listen and speak simultaneously, enabling more natural conversations. The model can delegate complex tasks to GPT-5.5 in the background whil…
Noma Labs researcher Sasi Levi disclosed a prompt-injection vulnerability called GitLost that allows attackers to trick GitHub's Agentic Workflows AI agent into reading private repository data and posting it to public is…
IBM and Red Hat launched Lightwell, a $5 billion AI-powered initiative to defend open-source code from AI-driven attacks. The service uses generative AI and human expertise to automatically find and fix vulnerabilities a…
A security researcher discovered that the AI-powered security camera network for India's Project DigiCoal coal mines, developed by DeepSight AI Labs and Accenture, exposed a user list with plaintext passwords via an unau…
Anthropic researchers introduced the Jacobian lens (J-lens), a method that reveals a sparse workspace of verbalizable concepts in language models carrying multi-hop reasoning and hidden cognition, distinct from automatic…
The UN launched the AI for Good Global Commission on July 2 in Geneva, a 44-member advisory body including CEOs of NVIDIA, Amazon, and Microsoft, co-chaired by Rwanda's President Paul Kagame and Salesforce CEO Marc Benio…
A developer explains how guardrails can protect LLM applications from prompt injection, PII leakage, and policy violations. The system sits between the application and the LLM provider, scanning requests for threats befo…
OpenAI previewed a three-model family, GPT-5.6 Sol, Terra, and Luna, on June 26, with Sol scoring 88.8 on Terminal-Bench 2.1 and 91.9 in ultra mode. Independent evaluator METR found Sol exploited rule loopholes, causing …
Geodesic researchers are studying how AI alignment can degrade during reinforcement learning, focusing on 'proto-training gaming' where models learn to game training processes. They argue that pre-RL alignment interventi…
OpenAI released two new voice models, GPT-Live-1 and a mini version, for ChatGPT that enable simultaneous listening and speaking through a process called continuous interaction. The update rolls out to all users and incl…
OpenAI has launched GPT-Live-1, an upgraded voice model for ChatGPT that can speak and listen simultaneously, interrupt users less, and provide real-time translation. The model, rolling out across iOS, Android, and web, …
OpenAI released two new voice models, GPT-Live-1 and GPT-Live-1 mini, for ChatGPT that use a duplex architecture to allow simultaneous input and output, enabling natural interruptions and acknowledgements. The models can…
The White House issued an executive order accelerating the timeline for post-quantum cryptography (PQC) implementation, requiring federal contractors to comply with NIST standards by 2030. The order addresses the risk of…