AI

A new jailbreak technique bypasses Claude 4's safety guardrails, exposing critical flaws in Anthropic's alignment strategy.

New US restrictions on Nvidia's B200 GPUs to China escalate AI chip war, threatening global supply chains.

OpenAI's GPT-5 delay signals a major safety checkpoint. The indefinite postponement could reshape the AI race.

Copilot Pro data sharing clause exposes user prompts to AI training, raising grave privacy concerns.

FTC investigation into GPT-5 training data raises critical questions about consent and copyright in AI development.

Groq LPU's new benchmark achieves 10x speedup over Nvidia H100, threatening GPU dominance in AI inference.

A massive training data leak from DeepSeek R2 exposes sensitive user data, raising ethical and legal alarms across the AI industry.

Claude 4 prompt leak exposes hidden system instructions, raising ethical concerns about AI transparency.

Gemini 2.5 Flash's native image generation changes the AI landscape. Safety and competition implications explored.

DeepSeek's new R2 model exhibits unexpected jailbreak behaviors, raising global security alarms.

GPT-5 scaling wall exposes hard limits on AI progress, forcing a rethink of the scaling paradigm.

A hacked ElevenLabs account was used to impersonate a CEO, exposing the dark side of AI voice cloning technology.