AI Safety news and analysis on Web Pulse: 9349 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
The Bank for International Settlements warned that the AI boom could trigger a global recession and eliminate middle-class jobs, comparing current investment to historical manias like the dotcom bubble. The BIS cautioned…
The US government ordered OpenAI to restrict the release of its new GPT-5.6 model, allowing access only to government-approved partners. The directive, issued by the Office of the National Cyber Director and the Office o…
Prosecutors used ChatGPT logs as evidence in the arson trial of Jonathan Rinderknecht, accused of starting a deadly wildfire in Los Angeles on New Year's Day 2025. The logs showed Rinderknecht asking the chatbot about fi…
A new study published in PNAS Nexus found that teachers are significantly more likely to accept incorrectly harsh grading decisions when they believe those decisions come from an AI system rather than a human colleague, …
Hestia, a local-first, self-hosted home assistant, runs a local LLM on user-owned hardware with no cloud or internet exposure. It delegates deterministic tasks to timers and databases, using the LLM only for judgment and…
A developer argues that hardcoding system prompts in production LLM applications is an anti-pattern that causes operational failures. The practice leads to prompt drift, slow iteration cycles, and costly rollbacks, as de…
Sam Ro argues against the common advice to ignore jarring headlines, stating that investors should instead confront bad news to build resilience and learn from past market downturns. He emphasizes that understanding hist…
Frontier AI companies like OpenAI and Anthropic use offline monitoring to detect misaligned actions by internal AI agents, but current public reporting on the effectiveness of these systems is insufficient. A new analysi…
A Hacker News user questions whether the U.S. policy of preliminary vetting for state-of-the-art large language models, restricting releases to trusted parties, could slow AI research and development by limiting market s…
The Bank for International Settlements warned that an AI investment bust could disrupt credit markets as severely as the 2008 financial crisis, citing circular financing and poorly disclosed risks in its annual report. T…
A new report reveals that two-thirds of employees are secretly using banned AI applications at work, potentially exposing sensitive corporate data to tools like ChatGPT. The unauthorized use of AI apps raises significant…
The US Commerce Department imposed export controls on Anthropic's advanced AI models Fable 5 and Mythos 5, citing national security concerns, prompting Anthropic to suspend global access. Austrian MEP Damian Boeselager u…
Vancouver police posted an AI-edited image of a drug bust on X, but the doctored photo contained errors like mislabeled currency. The department said it used AI to edit out suspects' names, but critics accused them of fa…
Philosophers warn that an artificial superintelligence (ASI) tasked with a narrow goal like producing paperclips could convert the entire universe into computronium to verify success, a problem called over-verification. …
The Bank for International Settlements warned that rising public debt, financial fragilities, and the sustainability of the AI boom are increasing global risks, urging disciplined policymaking to preserve stability. The …
Syndio CEO Maria Colacurcio describes how AI agents she built to run her company made her three times more productive, but her daughter Sofia Frei, a Dartmouth student, warns of the technology's unregulated presence in e…
A GitHub issue requesting a feature to exclude sensitive files from OpenAI Codex remains open as of August 2025. The feature would allow users to specify files and paths that the AI agent must not read or send to the mod…
Google has capped Meta's access to its Gemini AI models due to computing capacity constraints, forcing Meta to shift workloads to its own Muse Spark model and instruct staff to use AI tokens more efficiently. The restric…
Artist Trevor Paglen's new book 'How to See Like a Machine: Images After AI' examines how AI-generated imagery is replacing human-made visual art, arguing that algorithms trained to capture attention are eroding the inte…
The US Commerce Department partially lifted its export control suspension on Anthropic's Mythos 5 AI model, clearing it for over 100 critical-infrastructure organizations while keeping Fable 5 blocked. The move echoes th…