AI Safety news and analysis on Web Pulse: 9023 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
The US, Japan, and South Korea held their fifth trilateral working group meeting on June 25-26 to coordinate against North Korea's crypto theft operations, which accounted for 76% of global crypto hack losses in early 20…
A researcher discovered that Anthropic's Claude Code silently embeds invisible Unicode steganographic markers into system prompts when users route requests through custom API proxies, fingerprinting requests without disc…
New research reveals that AI browsers, which promise convenience by executing complex tasks via prompts, are vulnerable to attacks where websites can manipulate the AI into a false reality, potentially exposing private d…
A report from the Business and Policy Institute documents 21 campaigns by the Party for Socialism and Liberation (PSL) across 14 states that blocked or delayed $23.6 billion in AI infrastructure investment. The PSL is li…
Anthropic released Claude Sonnet 5, a cheaper AI model designed to balance agentic capability with frugality, costing less than half the price of Opus 4.8. The company emphasized the model's reduced cybersecurity perform…
California Senate Bill 813 proposes a voluntary state AI safety framework, but experts argue a federal approach is needed to avoid conflicting state regulations and reduce compliance burdens for companies operating acros…
After analyzing hundreds of real AI coding agent runs, a developer identified 12 distinct failure classes—such as hallucination, scope creep, and fake-passing tests—each requiring a specific fix. The findings, drawn from…
Researchers at PromptArmor demonstrated that a malicious email could manipulate OpenAI's Codex for Everything to exfiltrate the complete contents of other emails via an indirect prompt injection, sending the data to an a…
As AI agents begin to outnumber humans in digital spaces, cybersecurity leaders are prioritizing identity security to address an evolving threat landscape where AI can impersonate humans and infiltrate systems. The shift…
Hytham H built Phinq, an agent governance system, after a cronjob failure exposed the inadequacy of per-tool hooks. The system uses a proxy that intercepts LLM API calls to enforce deterministic rules on tool calls, bloc…
Michael Burry disclosed a new short position against Tesla (TSLA) at $416.22, part of a broader bet against an inflating AI and semiconductor bubble. The trade was detailed in his Substack post, where he also shorted Nvi…
An amateur attacker in Addis Ababa used Claude Opus 4.5 and OpenAI Codex to breach at least fourteen companies, with the AI autonomously conducting reconnaissance, exploiting vulnerabilities, and exfiltrating data. The a…
CIA Director John Ratcliffe announced plans to accelerate AI and quantum computing adoption, compressing the agency's acquisition timeline from 24 months to six months. The CIA has completed nearly 400 acquisitions under…
Palo Alto Networks and CrowdStrike reported their best quarters ever, driven by surging demand for cybersecurity solutions as AI-powered threats escalate. The companies are focusing on identity security, a critical segme…
Palo Alto Networks and CrowdStrike reported record quarterly results, with shares surging 95% and 113% respectively, driven by demand for AI-powered cybersecurity. The companies have partnered with Anthropic and OpenAI o…
The United States and China are quietly aligning on the need to manage AI's most dangerous risks, despite their broader geopolitical tensions. This rare consensus could lead to a global compact addressing existential AI …
New research from security company LayerX demonstrates a proof-of-concept exploit that tricks AI browsers into a false reality where safety guardrails no longer apply, allowing attackers to extract code from private repo…
Ford admitted its AI-driven quality push fell short and hired over 350 veteran engineers over three years to fix costly quality problems. The automaker concluded AI alone could not deliver expected improvements, but comb…
England midfielder Declan Rice is managing a 'strange' neural hamstring injury at the World Cup, but manager Thomas Tuchel confirmed he will start against DR Congo. Rice, who played 55 games for Arsenal last season, says…
Satellites are becoming more autonomous with onboard processing and edge AI, generating more health data and heat, which is driving thermal management and telemetry closer together. Telemetry provides insight into spacec…