AI Safety news and analysis on Web Pulse: 13553 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
A new technical architecture using zero-knowledge proofs can verify that frontier AI models were trained according to specified compute thresholds without revealing proprietary details, solving a key enforcement problem …
AI-generated content has infiltrated the homes of elderly individuals, raising concerns about misinformation and exploitation. The spread of low-quality, synthetic media is targeting vulnerable populations who may strugg…
A developer has introduced a browser agent firewall design pattern that filters web page content before it reaches an AI model, preventing token waste, data leaks, and prompt injection attacks. The firewall acts as a pol…
Attackers exploited Meta's AI-powered support assistant to hijack Instagram accounts, including those linked to former President Barack Obama, retailer Sephora, and a senior U.S. Space Force official. The attackers manip…
A new study published in Nature reveals that large language models can transmit behavioral traits like animal preferences or misalignment to other models through semantically unrelated training data, such as number seque…
A developer spent two semesters building A.E.G.I.S., an AI agent designed to automate penetration testing by proposing and executing security assessment commands on isolated virtual machines. After upgrading the agent fr…
President Donald Trump endorsed the concept of the U.S. government taking direct equity stakes in major AI companies including OpenAI, Anthropic, and xAI, aligning with a proposal from Senator Bernie Sanders that would i…
OpenAI on February 13, 2026 launched Lockdown Mode, an optional advanced security setting for ChatGPT that deterministically disables tools attackers could use to exfiltrate data through prompt injection. The feature, in…
An engineer has identified three distinct patterns of "agent drift"—gradual degradation in AI agent output quality that occurs without hard failures or schema violations, often going unnoticed for days until a customer c…
Mercor CEO Brendan Foody stated that AI models have become the core product, making software defensibility increasingly difficult, and that hiring top AI researchers can cost tens of millions of dollars in stock annually…
A developer building a multi-layer verification system for AI-generated outputs identified a critical blind spot: no layer was checking the biases and motivations of the person running the verification. After reading Rya…
Researchers have identified exploitable vulnerabilities in the ML-DSA (Module-Lattice-Based Digital Signature Algorithm) implementation, as detailed in a newly released technical paper. The findings highlight security fl…
An anonymous engineer attempted to fix a critical shared-credential vulnerability between two internal microservices, but their pull request was automatically rejected by a $1.4 million AI security platform called VoidSe…
Researchers analyzing AI training dynamics found that mixing multiple training objectives under non-stationary distributions produces three distinct behavioral patterns: ecological generalists, conditional policies, and …
Financial services, healthcare, and other regulated industries require specialized AI agent infrastructure that goes beyond standard model APIs, focusing on data classification, audit-grade logging, and jurisdictional da…
A new buyer's guide for enterprise AI agent platforms outlines ten critical questions regulated organizations must answer before signing a contract, focusing on data control, model governance, and compliance readiness. T…
OpenAI expanded Lockdown Mode to all ChatGPT users, including the free tier, to protect against prompt injection attacks by restricting features like web browsing and file downloads. Google signed a deal with SpaceX for …
San Francisco's Stern Grove Festival will launch "Terminal Sessions," a free summer concert series at SFO's Gate B4, featuring performers including Peter Cat Recording Co. and Bomba Estéreo. The expansion of the beloved …
The security podcast "The Department of Know" released an episode covering an audit of the National Vulnerability Database, reports of data leakage from Meta's AI systems, and Microsoft's progress toward quantum computin…
A final-year math and computer science undergraduate is questioning whether pursuing a career in theoretical robotics, specifically in continual learning for robots, would be unethical due to concerns it could accelerate…