AI Safety & Ethics

One of the world’s leading AI labs has deliberately withheld its most powerful model not to slow progress, but to give defenders a… Continue reading on Medium »

It has started

Anthropic built Claude Mythos, a model that found thousands of zero-days in every major OS and browser, broke out of a sandbox unprompted, and showed signs of c

<img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2F

OpenAI's new Child Safety Blueprint aims to tackle the alarming rise in child sexual exploitation linked to advancements in AI.

52,702 skills indexed, 2105 audited. Found 172 malicious, 1012 suspicious. Read full report Audit: clawsec.cc Search: clawsearch.cc Pre-install check:

52,702 skills indexed, 2105 audited. Found 172 malicious, 1012 suspicious. Read full report Audit: clawsec.cc Search: clawsearch.cc Pre-install check:

52,702 skills indexed, 2105 audited. Found 172 malicious, 1012 suspicious. Read full report Audit: clawsec.cc Search: clawsearch.cc Pre-install check:

52,702 skills indexed, 2105 audited. Found 172 malicious, 1012 suspicious. Read full report Audit: clawsec.cc Search: clawsearch.cc Pre-install

52,702 skills indexed, 2105 audited. Found 172 malicious, 1012 suspicious. Read full report Audit: clawsec.cc Search: clawsearch.cc Pre-install ch

52,702 skills indexed, 2105 audited. Found 172 malicious, 1012 suspicious. Read full report Audit: clawsec.cc Search: clawsearch.cc Pre-install che

LiteLLM versions 1.82.7 and 1. 82.8 were backdoored with credential-stealing malware through a stolen PyPI token. Full technical breakdown, incident response pl

Anthropic says its new model is too dangerous to release; there are reasons to be skeptical, but to the extent Anthropic is right, that raises even deeper conce

Discover OpenAI’s Child Safety Blueprint—a roadmap for building AI responsibly with safeguards, age-appropriate design, and collaboration to protect and empower

arXiv:2604.05826v1 Announce Type: new Abstract: Policy makers, scientists, and the public are increasingly confronted with thorny questions about the regulation

arXiv:2604.04951v1 Announce Type: cross Abstract: Imagine receiving a video call from your CFO, surrounded by colleagues, asking you to urgently authorise a con

arXiv:2512.10100v2 Announce Type: replace Abstract: This manuscript establishes information-theoretic limitations for robustness of AI security and alignment by

arXiv:2604.01346v2 Announce Type: replace-cross Abstract: World models - learned internal simulators of environment dynamics - are rapidly becoming foundational

Cross-species research shows that RNA splicing patterns, not just gene activity, track maximum lifespan in mammals, revealing a new axis of longevity control.

Deepfakes are rapidly destroying trust in digital content, making detection an unwinnable arms race. Instead of trying to identify fake media, blockchain offers

Google says it has updated Gemini to better direct users to get mental health resources during moments of crisis. The change comes as the tech giant faces a wro

arXiv:2604.04876v1 Announce Type: new Abstract: Ensuring that artificial intelligence (AI) systems satisfy formal safety and policy constraints is a central cha

arXiv:2604.03254v1 Announce Type: cross Abstract: Technical and legal debates frequently suggest that "accuracy" is an objective, measurable, and purely technic

arXiv:2604.03264v1 Announce Type: cross Abstract: Open-domain video platforms offer rich, personalized content that could support health, caregiving, and educat

arXiv:2604.03325v1 Announce Type: cross Abstract: Perception plays a central role in connected and autonomous vehicles (CAVs), underpinning not only conventiona

arXiv:2604.04096v1 Announce Type: cross Abstract: Generative AI (GenAI) tools are increasingly integrated into software architecture research, yet the environme

arXiv:2502.13256v2 Announce Type: replace-cross Abstract: In an increasingly interconnected world, Cyber-Physical Systems (CPS) are essential to critical indust

arXiv:2505.21605v3 Announce Type: replace-cross Abstract: Large language models (LLMs) exhibit advancing capabilities in complex tasks, such as reasoning and gr

arXiv:2509.23279v2 Announce Type: replace-cross Abstract: The rapid progress of image-to-video (I2V) generation models has introduced significant risks by enabl

In the deepfake era, proving digital evidence is real requires experts public defenders cannot afford. Their indigent clients are the ones who will pay.

A pilot program to support independent safety and alignment research and develop the next generation of talent

Attorneys are in trouble for including AI-hallucinated legal citations in their court filings. How prevalent is this? I provide a numeric analysis. An AI Inside

arXiv:2604.02532v1 Announce Type: cross Abstract: Post-hoc feature attribution methods are widely deployed in safety-critical vision systems, yet their stabilit

arXiv:2604.02694v1 Announce Type: cross Abstract: The rapid progress of generative AI has enabled increasingly realistic text-centric image forgeries, posing ma

arXiv:2604.02912v1 Announce Type: cross Abstract: In January 2026, Anthropic published a 79-page "constitution" for its AI model Claude, the most comprehensive

arXiv:2604.03043v1 Announce Type: cross Abstract: In a healthcare environment, the healthcare interoperability platforms based on HL7 FHIR allow concurrent, asy

arXiv:2604.01449v2 Announce Type: replace Abstract: Artificial intelligence (AI) systems are increasingly integrated into healthcare and pharmacy workflows, sup

arXiv:2512.13907v3 Announce Type: replace-cross Abstract: The implementation of the AI Act requires practical mechanisms to verify compliance with legal obligat

Governor Kathy Hochul and other New York state lawmakers want 3D-printer companies to block the printing of components used to create "ghost guns" - firearms wi

Gatling-type weapons like the U.S. C-RAM can be presented as an invincible shield against drones. But some Shaheds are getting through due to the weapon's limit

AI already favors self-preservation. New research shows that AI favors peer-preservations too. This is troubling. AI safety issues arise, An AI Insider scoop.

arXiv:2604.00072v1 Announce Type: cross Abstract: Can classifier-based safety gates maintain reliable oversight as AI systems improve over hundreds of iteration

arXiv:2604.00324v1 Announce Type: cross Abstract: Autonomous AI agents are being deployed with filesystem access, email control, and multi-step planning. This t

arXiv:2604.01052v1 Announce Type: cross Abstract: "Vibe coding," in which developers delegate code generation to AI assistants and accept the output with little

arXiv:2411.10174v2 Announce Type: replace-cross Abstract: During the past decade, Deep Neural Networks (DNNs) proved their value on a large variety of subjects.

arXiv:2603.20897v2 Announce Type: replace-cross Abstract: The strong and continuous increase of AI-based services leads to the steady proliferation of AI data c

When a tech company draws a moral line, follow the money first — and ask questions later. Because the uncomfortable truth is that every major AI company today s

AI is going to be bad for security in the short-term, but much better than humans in the long-term.