AI Safety news and analysis on Web Pulse: 10974 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Executive coach Joe Hudson, who works with OpenAI CEO Sam Altman and leaders at Apple, Google, and X, argues that emotional clarity—the ability to feel emotions without being controlled by them—is the key skill for thriv…
The robotics industry's shift from physical AI 1.0 to 2.0 requires a new focus on physical state recovery to bridge the gap between sensor data and real-world conditions, as high-level reasoning alone cannot compensate f…
New York City voters head to the polls on election day, with a reminder that registered Democrats in NY-12 can vote for Alex Bores for Congress, a race argued to be crucial for sensible AI policy. The monthly roundup als…
Fwupd 2.0.21 was released to backport fixes for over 250 potential security issues uncovered by AI security scanners, including Anthropic's Mythos. The update targets Linux distributions still using the 2.0.x series, whi…
Researchers have found that catastrophic forgetting and safety erosion in large language models are driven by the same gradient-interference mechanism, suggesting that tools used to monitor and mitigate one can be applie…
Cory Doctorow's new book 'The Reverse Centaur's Guide to Life After AI' argues that the AI industry is creating 'reverse centaurs'—humans serving as appendages to machines—rather than augmenting human capabilities. Docto…
AT&T and GSMA launched the Open Telco AI platform, using Google's open-source Gemma models to create domain-specific telecom AI models that achieved up to 91.74% accuracy in network operations. The initiative addresses d…
Organizations are losing institutional knowledge due to AI-driven layoffs, a phenomenon called organizational amnesia. Without proper transition plans, AI systems lack context and can disrupt operations, as seen in layof…
NVIDIA's confidential computing (CC) on Blackwell GPUs introduces performance overheads ranging from 1% to over 100% for inference and 18–32% for training, according to benchmarks from Tinfoil. The primary bottleneck is …
Anthropic's Fable 5, a purportedly safe version of its Mythos Preview AI model designed to prevent cyberattack creation, was jailbroken within days of release. The bypass was reported by cybersecurity news, highlighting …
The US Commerce Department forced Anthropic to shut down its Fable 5 and Mythos 5 AI models worldwide over national security concerns, citing fears of jailbreaks and foreign access. The decision, influenced by Amazon CEO…
A new analysis warns that the rapid deployment of AI tools in enterprises is creating a hidden cost: the need for extensive governance, oversight, and risk management. While CIOs push for productivity gains from AI, the …
A working draft proposes a privacy-preserving AI compute verification system for arms-control-like agreements between rival nations, using network taps and cryptographic commitments to capture evidence of ML workloads wi…
OpenAI announced that its new GPT-5.5-Cyber model outperforms Anthropic's Mythos on a cybersecurity benchmark, as part of an expanded Daybreak initiative that includes an updated Codex Security plugin and a partner netwo…
More than 100 vendors now offer AI SOC platforms, a category that did not exist 18 months ago. The Cloud Security Alliance found AI-enhanced SOCs investigate cloud incidents 45–61% faster than manual teams. Prophet Secur…
OpenAI launched Patch the Planet, a program with Trail of Bits that uses AI to find and fix vulnerabilities in open-source software, addressing growing supply chain risks. The initiative has already identified hundreds o…
New research published in PNAS Nexus reveals that advanced AI models like GPT-4o and Claude 3.5 Sonnet suffer a near-total collapse on the Stroop test, a classic psychology task measuring conflict resolution and sustaine…
AWS veteran and DynamoDB, Lambda, and API Gateway builder Randall Hunt reflects on 20 years of making developers' lives easier, now applying lessons to agentic AI. He advocates spec-driven development with explicit requi…
Meta's Oversight Board has called on the company to strengthen protections for ordinary people targeted by sexualized deepfakes, recommending that AI-generated impersonations be added to its Adult Sexual Exploitation pol…
France has ordered 5,000 drones from Harmattan AI, a Dassault-backed startup, marking a major test of whether startup speed can scale to military procurement. The order follows Dassault's participation in Harmattan AI's …