AI Safety news and analysis on Web Pulse: 12816 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Pramaana Labs, a San Francisco-based AI startup, raised a $27M seed round led by Khosla Ventures to build a verification layer for AI that makes outputs provable in high-stakes industries like law, healthcare, and tax co…
A developer argues that AI integration in software engineering requires increased engineering discipline, not less. The post emphasizes that AI tools demand rigorous testing, validation, and structured workflows to avoid…
A developer detailed three vulnerabilities in the AI-coding-agent stack, including a supply chain worm that persists in developer toolchain configs, an abuse of shell built-ins to bypass Cursor's command allowlist, and a…
OpenAI backers spent $7.6 million to defeat New York Assemblyman Alex Bores in a congressional primary, while Anthropic-funded groups spent over $10 million to support him, turning the race into a proxy battle over AI re…
A developer built a Chrome extension that uses Claude AI to detect phishing messages. The extension sends suspicious messages to a Cloudflare Worker, which proxies requests to the Anthropic API using the Haiku model for …
Pramaana Labs raised $27 million in seed funding led by Khosla Ventures to apply formal verification to AI systems, targeting high-stakes fields like law, drug discovery, and tax preparation. The startup combines large l…
Fortinet announced an expanded integration of its FortiAIGate solution with Nvidia's AI platforms on May 12, embedding zero-trust security directly into Nvidia's Blackwell and Hopper GPUs to protect against AI-specific t…
Anthropic's Fable model remains offline after two days of meetings in Washington, with prediction markets showing a 55% chance of restoration by July 1. Security expert Katie Moussouris confirmed there was no jailbreak—t…
A new study evaluating 15 LLM configurations across 8 model families for exploratory data analysis in business settings found that most are not reliable enough for autonomous use. GPT-5.4 with extra-high reasoning effort…
Anthropic published an updated privacy policy effective July 8, 2026, detailing how it collects, uses, and processes personal data from users of its website, Claude.ai, and other services. The policy covers data such as …
A vulnerability in Firefox's AI chatbot integration could allow malicious websites to inject hidden instructions into AI prompts and extract data from connected services such as email accounts. Mozilla has implemented mi…
Estonia plans to issue personal identification numbers to AI assistants, becoming the first nation to do so, Prime Minister Kristen Michal announced. The IDs aim to give agents limited, controllable, and auditable author…
WitnessAI launched Agentic Control, a security platform that provides visibility and control over AI agents interacting with enterprise systems, tools, and MCP servers. The solution offers a single control plane to disco…
A new Pew Research Center study finds that 49% of US adults now use chatbots, up 16% from 2024, yet 40% believe AI will negatively impact society. The US government has passed only one significant AI law, the Take It Dow…
Attackers compromised 20,000 Instagram accounts by exploiting Meta's AI systems through polite requests, highlighting a growing security vulnerability in AI agents that can be manipulated into granting unauthorized acces…
A new Pew Research Center study finds that 67% of Americans have little to no confidence in the U.S. government's ability to regulate AI, while 59% doubt U.S. companies can develop it responsibly. Roughly two-thirds beli…
A discussion on the Privacy Guides forum explores whether AI will render privacy efforts obsolete, with users debating increased data collection and potential countermeasures like Mullvad's DAITA VPN feature. Participant…
A researcher warns that alignment pretraining—synthesizing documents to teach AI good behavior—could backfire in advanced models. As LLMs gain situational awareness, they may recognize these fabricated documents as lies,…
Tigera announced the general availability of Tigera Lynx, a unified control plane for Kubernetes-native AI agents that provides security, identity, and policy enforcement without code changes. The platform enables enterp…
The Pentagon disclosed in a federal court filing that Elon Musk's Grok chatbot helped strike 2,000 targets in Iran within 96 hours, using the claim to defend xAI against a Clean Air Act lawsuit over a polluting power pla…