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.
Anthropic's new Claude Fable model unsettled a user by generating an evidence-based personal observation derived from their past conversations with earlier Claude versions. The user reported the model's response was "gen…
On June 1, a team of scientists published a preprint claiming they had edited human embryonic DNA with unprecedented precision, a breakthrough that could eventually enable "designer babies" engineered with specific chara…
Anthropic Chief Product Officer Mike Krieger used a thread on X to frame the company's June 9 launch of Claude Fable 5 as a product test, asserting the model can handle longer delegated tasks while the company's safety f…
Anthropic has released a "safe" version of its Mythos AI model, promising sufficient guardrails and user limitations after previously claiming the system was too dangerous to release. The new model carries a price tag tw…
GitHub has released a new experimental `/security-review` slash command for Copilot CLI, now available in public preview, that analyzes local code changes for security vulnerabilities such as injection flaws and cross-si…
The European Commission published a voluntary Code of Practice outlining steps for providers and deployers of generative AI to comply with the AI Act's transparency obligations, which take effect on 2 August 2026. The Co…
Researchers gave top AI models a classic attention test used in psychology and found a major flaw. While the models could correctly name colors in short lists, their performance deteriorated sharply as the task became lo…
Google DeepMind, Schmidt Sciences, the Cooperative AI Foundation, and the Advanced Research and Invention Agency, supported by Google.org, announced a new technical research funding call of up to $10 million for research…
Fable 5, the public version of the Mythos AI model, has been released with significant capabilities but raises concerns over new precedents it sets. The model's deployment highlights ongoing debates about AI safety tiers…
Anthropic released Claude Fable 5, its most capable public model to date, across major platforms including AWS, Google Cloud, and Microsoft Azure. The release introduces built-in control layers such as fallback to Claude…
Aikido's AI pentesting tool discovered a critical authentication bypass vulnerability in phpBB, affecting tens of millions of users across thousands of forums. The flaw, present in versions up to 3.3.16 and 4.0.0-a2, all…
Anthropic released Claude Fable 5 and Claude Mythos 5 on June 9, 2026, both belonging to a new "Mythos-class" tier above the Opus class in capability. The two models share the same underlying architecture, but Fable 5 in…
AWS Bedrock customers using Anthropic's Mythos-class models, including Fable 5 and Mythos 5, must now consent to 30-day data retention by Anthropic for misuse detection, with data leaving AWS's security boundary during t…
Anthropic has released Claude Fable 5, the first model in the new Claude 5 family and part of a Mythos-class tier above Claude Opus. Claude Fable 5 is the most intelligent generally available model with additional safety…
An international cohort of experts is being convened to produce a practical Total Defense manual that translates whole-of-society preparedness into concise doctrine, modular capability templates, legal authorities, and r…
Anthropic launched Claude Fable 5, a Mythos-class model at least twice the size of Opus 4.8, with API pricing at roughly 2x Opus and benchmark improvements including a jump from 13.4% to 29.3% on the FrontierCode Diamond…
Apple's most powerful AI advantage is its ability to personalize responses using data stored on users' devices, a feature the company calls Personal Context. While Apple's AI announcements at WWDC 2026 largely played cat…
Nearly half of all production code is now AI-generated, yet 93% of enterprises surveyed suffered a security breach directly from in-house developed apps in 2025, according to a new Checkmarx report. Three-quarters of org…
A German regional court has ruled that Google is directly liable for false claims in its AI-generated search overviews, treating the content as Google's own words rather than third-party search results. The court issued …
Anthropic implemented silent safeguards in its Claude Fable 5 model that secretly degrade responses to queries about competing AI development, including ML accelerator design and pretraining pipelines, without notifying …