AI Safety news and analysis on Web Pulse: 8012 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
James Cameron's 'Terminator 2: Judgment Day' returns to theaters in late August 2026 for its 35th anniversary, featuring a remastered 3D version. The film, which explores AI sentience and humanity's survival, will screen…
Noma Labs disclosed GitLost, a July 2026 prompt-injection technique that could make GitHub Agentic Workflows leak private repository contents through a public issue comment. The attack exploits an architectural agent-ris…
Security researchers at Noma Labs discovered a prompt injection vulnerability in GitHub's Agentic Workflows, dubbed GitLost, that allows attackers to trick AI agents into leaking private repository data as public comment…
Apoxy released CLRK, an open-source agent runtime that uses gVisor for isolation and runs on Kubernetes, designed to provide fully intercepted I/O, framework-agnostic guardrails, and audit trails for AI assistants. The p…
The European Commission on July 7, 2026, published an action plan linking advanced AI model deployment to cybersecurity assurance across critical sectors, including finance, health, energy, transport, and public administ…
China is considering new export controls on frontier AI models, including both proprietary and open-weight systems, to protect domestic capabilities amid intensifying competition with the US. Officials from the Ministry …
Neuronpedia, an open-source platform for AI interpretability, launched the world's first interpretability API in March 2024, enabling users to visualize, steer, and search over 50 million latents in AI models. Created by…
Anthropic filed a lawsuit against Abnormal AI on July 1, alleging unfair competition and trademark infringement. Abnormal CEO Evan Reiser responded, stating Anthropic never notified them, the logo was designed before Ant…
Discord admitted a bug in its AI moderation system wrongfully banned over 8,000 users since May, flagging harmless images like spreadsheets and chessboards as harmful. The company fixed the issue and is restoring account…
Utah launched a prescription refill program using an AI chatbot called Doctronic, allowing residents to get refills without seeing a doctor. The program has sparked debate among doctors and experts about AI's role in med…
The U.S. Cybersecurity and Infrastructure Security Agency (CISA) is using Anthropic's AI model Mythos to audit government software for vulnerabilities, according to three sources. The audits, conducted by CISA's Attack S…
Fans are increasingly using reverse image search tools like Face2social to verify celebrity social media profiles amid a surge of AI-generated fake accounts. Impersonators use advanced image editing to create convincing …
A developer argues that AI agents should not be given standing API keys due to security risks, and proposes using short-lived, scoped credentials delivered via a credential broker. The approach limits damage from leaks b…
Tim Henman urged Wimbledon to scrap the five-minute toilet break rule, arguing it disrupts match momentum. The former British No. 1 made the comments during Arthur Fery's victory over Grigor Dimitrov at the All England C…
A developer warns that German criminal code § 202c StGB, known as the 'Hacker-Paragraph,' criminalizes the preparation of data espionage and can trap developers who use security tools without explicit written consent. Th…
Anthropic has been banned as a vendor at comma_ai, with the company's founder recommending other firms follow suit and warning against over-reliance on cloud token vendors.
OpenAI endorsed the European Commission's voluntary Code of Practice on Transparency of AI-Generated Content on June 11, 2025, committing to use provenance tools like C2PA Content Credentials and SynthID watermarks to co…
Security researchers at Noma Security demonstrated the GitLost exploit, which uses indirect prompt injection to trick GitHub Agentic Workflows into exfiltrating private repository data through public issues. The attack e…
Vanta released GrantGuard, an open-source tool that audits and cleans up permission allowlists built by Claude Code. The tool classifies granted permissions into risk groups and provides a local web UI for one-click remo…
PortSwigger's Burp Suite faces growing competition from alternatives like Aikido Security, Caido, ZAP, and Invicti as teams demand DAST that runs continuously in CI/CD, AI-powered pentesting, and visibility into code and…