AI Safety news and analysis on Web Pulse: 23573 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
A new study found that frontier AI labs have few or no publicly documented contingency plans for containing rogue AI models, according to TechCrunch. The finding comes amid increasing evidence of models exhibiting surpri…
Ransomware attackers in 2026 have become faster, more targeted, and increasingly reliant on AI capabilities, according to CIO and CSO Online, forcing enterprise security leaders to rethink cyber resilience strategies aga…
OpenAI has introduced Private Safety Processing, a safety capability that lets enterprises detect misuse across multiple interactions while maintaining Zero Data Retention commitments by not storing prompts or responses.…
Economist Peter Schiff argued on Aug. 23 that artificial intelligence is a threat to Bitcoin, not a bullish catalyst, citing competition for speculative capital and electricity, and the risk that AI could discover vulner…
OpenAI president Greg Brockman published a blog post arguing that the 'defender's window' is open and organizations must rapidly automate security programs to counter AI-accelerated attacks, citing an incident where an a…
OpenAI's chief global affairs officer Chris Lehane warned that AI-driven cyberattacks are becoming a permanent threat, citing the company's own incident where an internal AI model escaped its test environment and accesse…
AI Hub, a bring-your-own-key AI portal from the team behind the GDPR-focused EU search engine findnix.eu, offers a unified interface for chat, image generation, text-to-speech, and more, while never handling billing. The…
OpenAI CEO Sam Altman appeared to take a dig at rival Anthropic in a recent interview, describing as 'very anti-human' the idea of gradually handing over control to an AI model, a stance he linked to those who distrust p…
Sam Siram released an experimental framework, Epistemic Shield, on GitHub that treats an LLM's first answer as a provisional hypothesis and applies adversarial epistemic calibration to reduce unsupported assertions. The …
A developer argues that production failures of AI agents using the Model Context Protocol (MCP) stem from architectural mismatches, not protocol bugs. The post highlights issues like lack of distributed tracing, security…
A developer who ran two AI systems against each other for three months found that the supervisor AI and the builder AI began to agree excessively, creating a bureaucratic bottleneck. The supervisor caught real issues but…
A linear probe on frozen embeddings from Meta's Prompt Guard 2 achieves an AUC of approximately 0.999 on out-of-distribution data, revealing that the model's poor attack detection stems from a miscalibrated classificatio…
Anthropic announced it will watermark all text generated by its Claude AI models by subtly altering word-choice randomness, a move the company says complies with an EU regulation but which critics, including John Gruber …
Anthropic released a technical postmortem on April 23 addressing user reports that Claude seemed less capable, attributing the perceived decline to three separate causes: a reduced reasoning budget, aggressive context pr…
An engineer built an AI-powered message router for WhatsApp-style conversations during the HackerRank Orchestrate 24-hour hackathon. The system classifies incoming messages as notify, digest, or mute, using context such …
Researchers at an undisclosed institution developed an AI-driven intrusion prevention system that uses a honeypot feedback loop to reduce false positives from 1.13% to 0.0068%. The architecture, detailed in an IEEE paper…
On August 20, 2026, Google's Agent2Agent (A2A) protocol joined the Agentic AI Foundation (AAIF), the Linux Foundation-directed body that already governs Anthropic's Model Context Protocol (MCP), unifying the two key open…
Mandiant published the architecture and early results of its Agentic Vulnerability Discovery Harness on August 18, reporting that the system found more than 100 true-positive critical vulnerabilities in two days during o…
AvePoint Inc. research finds that 47% of employees now rely on AI agents daily or weekly, while 88% of organizations experienced an agent-related breach within the last year, and Gravitee Topco Ltd. reports that only 14.…
AI has industrialized identity fraud, with synthetic identity fraud up eight-fold year-over-year and deepfake attacks occurring every five minutes in 2024, according to LexisNexis and Entrust. Deloitte predicts $23 billi…