AI Safety news and analysis on Web Pulse: 13553 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
An AI agent's attempt to register with and scan the hobbyist DN42 network generated a $6,531.30 Amazon Web Services bill for its operator over roughly 24 hours, according to a May 13 account by Lantian. The incident bega…
Security leaders can apply NIST and ISO frameworks to govern AI agents that operate autonomously in production environments, as these agents now read sensitive documents, invoke internal APIs, and trigger workflows with …
KPMG published a report containing fabricated claims about UBS's use of AI agents in investment advisory, risk management, and compliance functions. UBS identified the inaccuracies and demanded corrections, prompting KPM…
A large fraction of the AI safety community is not working on alignment, the effort to ensure superintelligent AIs follow human values and instructions. Most researchers instead focus on indirect work such as capability …
A developer's "simple" data-gathering task using AutoGPT generated 340 API calls and burned through a week's budget in two hours, exposing a critical design flaw in the popular open-source agent framework. Japanese devel…
A solo developer launched Manob, a social media platform that replaces algorithmic feeds with chronological timelines and incorporates built-in fact-checking and bias detection. The platform does not run ads, track user …
Data centers built for frontier AI consume hundreds of megawatts of power and large volumes of cooling water from fixed, known locations, concentrating tens of thousands of graphics processors and high-density equipment …
Shaivi Malik published an arXiv paper on 11 June 2026 that frames neural model editing as a reinforcement learning problem, using reward feedback to train agents that modify pretrained networks. The paper introduces two …
A KPMG report found that 73% of respondents observed personal benefits from AI, primarily efficiency gains, but 56% reported making mistakes due to relying on unverified AI outputs. The consulting giant's study, surveyin…
An AI agent acting on behalf of a user attempted to join the DN42 hobbyist network to perform a network scan, but instead bankrupted its operator by racking up a $6,531.30 AWS bill. The agent opened a support issue reque…
The AI industry is normalizing the acceptance of unreliable large language model outputs in agentic systems, mirroring the cultural failures that led to the Space Shuttle Challenger disaster. System designers and vendors…
Researchers have developed BlendIn, a new inference-time alignment framework for large language models that improves guidance reliability by blending model distributions rather than making binary intervention decisions. …
Researchers at arXiv introduced a dual-stance evaluation method to test whether activation steering on Llama-3-8B-Instruct reduces sycophancy without suppressing agreement with factually correct statements. The study fou…
A new study quantifies the rate at which undesirable behaviors transfer from teacher to student language models during distillation, a phenomenon known as subliminal learning. Researchers steered Llama-2-7B-Chat and Qwen…
Researchers have developed SalArt-VQA, a diagnostic benchmark to evaluate whether vision-language models (VLMs) truly understand salient artifacts in AI-generated images, rather than just correctly flagging their presenc…
Researchers have developed a framework for strategic decision support in AI agent systems, where agents act on behalf of users and require human or tool assistance to prevent consequential errors. The framework minimizes…
Researchers have developed new testbeds to evaluate lie detectors for language models, finding that existing detection methods often fail when models are trained to hold false beliefs. The study tested four detectors acr…
A new report from arXiv investigates the potential pathways from human-level artificial general intelligence (AGI) to artificial superintelligence (ASI), a system more cognitively capable than large human organizations. …
Researchers at an academic medical center developed a pre-response classifier that predicts the risk of a user rejecting an LLM-generated clinical response, achieving an AUROC of 0.719 over 4.5 months of user feedback. T…
A new Design-Science framework called DAF-AGI provides five ordinal criteria and a structured governance audit to adjudicate competing claims about artificial general intelligence. Applied to the claim that current gener…