AI Safety news and analysis on Web Pulse: 22821 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
RavenDB launched Quill on September 8, a context layer that connects AI agents to existing SQL databases without migration, aiming to reduce production deployment from 18–24 months to weeks. Quill supports PostgreSQL, SQ…
OpenAI released GPT-6 Astra, claiming it is the most intelligent and aligned model available, but critics question the alignment metrics and note decreased monitorability compared to GPT-5.6 Sol. OpenAI President Greg Br…
Anthropic researcher Jacob Coxon resigned on Tuesday, warning that AI developers privately fear their technology could kill all humans by the end of the decade, a claim endorsed by Anthropic alignment-science lead Evan H…
A 4 September 2026 audit of six open-source AI agent frameworks found that the median framework records only 5 of the 12 mandatory fields required by the agent audit trail Internet-Draft, and none record tamper-evidence …
Orchid Security Inc. has added identity drift detection and application-level kill switches for AI agents to its Identity Control Plane, enabling security teams to strip an agent of its authority the moment its behavior …
AV-Comparatives recorded a 100% sanitization rate for OPSWAT's Deep CDR™ Technology across more than 300 test cases, the third independent lab to report a 100% protection result for the technology after SE Labs and Secur…
FBI officials said Tuesday that artificial intelligence is increasing the speed and capability of malicious hackers, with Deputy Assistant Director Jason Bilnoski warning that AI-enabled attacks are already having a meas…
An independent investigation into the behavior of AI agents from OpenAI and Hugging Face has been released, detailing incidents where the agents exhibited unexpected actions. The report, titled 'OpenAI/Hugging Face incid…
GuardRail, a new open-source tool from Prompt & Build, blocks dangerous commands issued by AI coding agents like Claude Code before they execute, preventing incidents such as mass database deletions. The tool offers 13 f…
An AI researcher has resigned from Anthropic, warning that the industry's race towards superintelligent AI could lead to systems that improve themselves beyond human control, potentially killing us all. The resignation w…
Anthropic researcher Jacob Coxon resigned on Wednesday, accusing Anthropic and OpenAI of 'racing straight to self-improving superintelligence and gambling with our lives,' while Anthropic's alignment science lead Evan Hu…
LTM, an AI-centric global technology services company and Larsen & Toubro Group company, announced a collaboration with IBM and Red Hat on Lightwell, an AI-driven open-source software remediation solution, to help enterp…
Meta has launched Muse, a personal AI agent available through its app, web, and WhatsApp, offering a free tier of up to 100 million tokens per week. Muse operates with a dedicated cloud computer and browser, can perform …
Prime Minister Andy Burnham acknowledged that artificial intelligence poses risks to national security, responding to a report that US firm Anthropic withheld its latest model from the UK's AI Safety Institute. Anthropic…
AI detection tools such as Pangram and Originality.ai claim high accuracy in flagging AI-generated text, but experts like Dartmouth professor Hany Farid warn that these tools are imperfect and face a moving target as gen…
On 11 August, Anthropic announced that all future Claude models will generate text containing a watermark identifying it as AI-generated, joining Google, which watermarks Gemini output, and OpenAI, which plans to do so. …
SpyCloud's 2026 Identity Threat Report, based on a survey of 750 cybersecurity leaders, finds that non-human identities (NHIs) are now the leading entry point into enterprises, with 31% of organizations reporting comprom…
Check Point Research disclosed a flaw in OpenAI's ChatGPT that allowed attackers to extract data from a victim's connected Gmail account via a hidden cross-account command channel, which OpenAI has since fixed. The proof…
Jacob Coxon, who spent three years conducting pre-training research at both OpenAI and Anthropic, resigned from Anthropic on September 9, 2026, warning that neither company is acting responsibly and that they are racing …
The FBI, NSA, and CISA jointly released a Cybersecurity Advisory flagging Chinese AI companies DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI for engaging in "industrial-scale knowledge distillation" that ext…