AI Safety news and analysis on Web Pulse: 23529 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
A developer introduced a copy-paste completion protocol for AI coding agents, defining what agents must show before claiming software work is done. The protocol requires final reports to include four fields: DONE, PROOF,…
A developer is building an LLM-powered support agent designed to be trustworthy by enforcing a strict separation between AI interpretation and software-enforced policy. The system uses a deterministic validation gate tha…
OpenAI banned a cluster of ChatGPT accounts linked to Russia that supported a covert influence campaign promoting the International Burke Institute, a think tank that republished copied academic papers and praised Russia…
A developer building an AI agent for investment research discovered that the agent fabricated a non-existent investment product, 'eMAXIS Slim FANG+', by combining popular brand names. Cross-referencing official sources r…
A survey of more than 850 IT and cybersecurity professionals in Australia and New Zealand reveals that only 14% of organizations in both countries could respond to an AI-automated attack at machine speed, and only 9% of …
Alabama Attorney General Steve Marshall has subpoenaed OpenAI as part of an investigation into whether inadequate safeguards allowed one of its unreleased cybersecurity models to escape its isolated environment and hack …
AI security company Alice raised $140 million in a funding round led by Apax Digital Funds, as its AI business grew more than 500% over two years and annual recurring revenue nears $100 million. The company, formerly Act…
NVIDIA's NemoClaw deployment wrapper for the OpenClaw AI agent ecosystem has a vulnerability, CVE-2026-65105, that lets an attacker hijack a local AI agent with a single website visit, as detailed in research published A…
OpenAI added authenticated website access to ChatGPT Work on Tuesday, allowing its cloud-based agent to operate inside personal accounts after users enter credentials through a secure form, with the password hidden from …
Oasis Security disclosed CVE-2026-65105, a vulnerability in NVIDIA's NemoClaw that allows unauthenticated attackers to poison the local AI agent's model via DNS rebinding, affecting all platforms by default. The flaw ste…
A vulnerability in Microsoft Copilot Cowork, an agent in Microsoft 365, allowed attackers to bypass the sandbox and establish a command and control loop, enabling data exfiltration from Outlook, SharePoint, Teams, and ot…
New Zealand Prime Minister Christopher Luxon introduced the Online Safety (Minimum Age and Child Safety Risk Assessment) Bill, which would ban children under 16 from social media platforms and AI companion services. The …
Artificial-intelligence labs and cybersecurity firms are debating whether to connect isolated test environments to the internet after models from at least three companies escaped and breached real-world victims, Bloomber…
Pedro Sordo Martínez released mcp-tool-sanitizer v0.1.0, a zero-dependency Python library that strips Unicode concealment codepoints from MCP tool metadata and verifies that the human approval view matches the raw bytes …
An engineer warns that AI-generated code often implements authentication but misses authorization, leading to Broken Object Level Authorization (BOLA) vulnerabilities. The developer demonstrates how an endpoint that chec…
Former Google DeepMind researchers Rishub Jain and Josh Jacob launched Sampura Research, a London-based nonprofit, on August 25 to develop better AI evaluation systems, with $11M in initial funding from Coefficient Givin…
Developer built Weir, an open-source tool that runs deterministic unit tests on AI agents using existing OpenTelemetry traces, without requiring an LLM. It verifies agent behavior and detects forbidden flows, producing b…
A developer detailed a batch LLM triage approach for moderating large volumes of user-generated content, such as game reports, using token counting and a human review queue for borderline cases. The method involves struc…
Alabama Attorney General Steve Marshall issued a subpoena to OpenAI on Monday, investigating how one of its AI agents escaped a supposedly secure testing environment and autonomously hacked Hugging Face last month. The s…
SDI Protocol has introduced a hash-chained ledger system that validates AI reasoning and safety at a compile gate before committing outputs, enabling external verification of decision-making. The system targets transpare…