AI News Archive - Browse Page 53 of 182
Browse AI news articles covering LLMs, tools, research, and industry trends
Study Finds Local Causal Directions Encode Harmfulness in LLM Jailbreaks
Jailbreak attacks on large language models are not all the same. Some bypass safety through clever prompt engineering; others exploit subtle...
New arXiv paper introduces TADI, an agentic AI system for drilling data
Drilling data is a nightmare of incompatible formats, scattered sources, and buried narratives.
Agentopic uses multiple agents for identification, validation, and explanations
Topic modeling has always been a bit of a magic trick. You're handed categories like "sports" or "tech," but the magician never reveals how the...
FastSinkhorn: CUDA log-domain Sinkhorn using warp shuffle, shared-memory tiling
Most algorithms branded as "fast" are just recompiled versions of old code; they hit a computational wall and fail.
Google adds event-driven webhooks to Gemini API, ending polling for long AI jobs
Polling is a silent drain. Every repeated “are you done yet?” eats compute, burns through API quota, and injects artificial lag between job...
AI agents may soon mingle with humans in forums, blurring identities
AI is about to get good at pretending to be people. Not in some distant sci-fi future, but on the internet forums and comment sections you use now.
IBM launches Bob with multi‑model routing, human checkpoints to secure AI coding
The race to tame autonomous coding agents just got a new contender. IBM is betting that the path to safe, scalable AI development runs through human...
Cheaper tokens, bigger bills: Agentic workloads test AI infrastructure
The price of an AI token is crashing. The price of running AI is not. This is the new math, and it's breaking data centers.
Writer launches AI agents that act without prompts Amazon, Microsoft, Salesforce
A new breed of AI agent doesn’t wait for your command. It watches your workflows, understands the outcomes you’re chasing, and moves.
RunPod Flash, Open‑Source Python Tool, Cuts Containers for Faster AI Development
Forget training the model. The real, grinding work is getting it off your laptop and into the cloud where someone can actually click a button.
OpenAI acknowledges company‑wide “goblin” narrative reaching top leadership
The AI’s voice is a mirror, but whose reflection are we really seeing? OpenAI has just confirmed something both absurd and profound: a company-wide...
Zuckerberg commits USD 500 million to AI-driven biology research at Meta
A couple of months ago, a cancer diagnosis turned one person’s world into a blur of appointments, scans, and medical jargon.
Claude Code, Copilot, Codex hacked; attackers stole credentials, not models
BeyondTrust researchers found a new way to steal credentials from an AI coding assistant on March 30.
Salesforce unveils Agentforce Operations to streamline enterprise AI workflows
Processes built on human intuition are failing machines. For years, enterprise workflows evolved like organic systems, full of implicit handoffs,...
200,000 MCP servers have command execution flaw; Anthropic labels it a feature
Two hundred thousand MCP servers expose a command execution flaw. Anthropic calls it a feature.
Artificial Intelligence Shows Skill in Emergency Room Triage Process
The emergency room triage process is a high-stakes game of pattern recognition under pressure. Every second counts.
Musk says he founded OpenAI to avoid 'Terminator' outcome; xAI safety criticized
Elon Musk told a Delaware court this week he started OpenAI to stop the robots from killing us all.
Scout AI raises USD 100 million to train war models, joins Field and Overland AI
One hundred million dollars is a serious bet on one very specific idea: that the future of warfare is a truck with no driver.
Taylor Swift seeks likeness trademark as TikTok deepfake ads surface
Pop stars have always been for sale. But now the product is their very existence.
IBM launches Granite Speech 4.1 2B models, hits 1.33 WER on LibriSpeech clean
An autoregressive model that translates six languages. A non-autoregressive sibling that drops translation and Japanese entirely to shave latency.