Latest AI and tech news
The story everyone covered was that Gemini hacked three firms. The story nobody covered is what that means for how frontier AI safety…...
“Fleet management vendors” fall into two layers. Telematics and safety vendors are built around the vehicle: GPS tracking, diagnostics, ELD and hours-of-service compliance, dashcams, and driver safety scoring. Dispatch and allocation platforms are bu...
Epoch AI researchers say the widely available AI model GLM 5.3 turned public security fixes into working exploits before updates reached Firefox users. It autonomously compromised six Firefox releases for about $1,000 in API fees each without trigger...
An Architect’s Guide to Zero Trust, Task Scoped Tokens, and Last Mile Enforcement...
As AI takes on more high-stakes financial processes, from payment approvals to pension fund transfers, how we deploy it becomes just as important as what we automate.
Baseten has released joey-1, a 2B model from its in-house research lab, Base Labs. The joey-1 decision model returns a label and a calibrated probability instead of generated text. It targets high-volume routing, classification and guardrail checks. ...
Anthropic gave 50 experts an AI agent whose preferences were set by coin toss. The experts followed the agent's view up to 86 percent of the time on judgment calls, and most believed the conclusions were their own....
An autonomous agent will do exactly what its tools let it do, and no more. This sounds like a constraint. In practice it is the whole risk, because the set of things an agent’s tools let it do is almost never the set anyone deliberately chose. It is ...
Prompt injection is currently the top risk when it comes to building applications driven by large language models (LLMs), like AI agents. That’s because prompts are the main control interface for these models, and they can follow text instructions em...
The post Luke Muehlhauser on the most important AI safety work nobody is doing appeared first on 80,000 Hours....
OpenAI’s security tests revealed a hidden communication channel between isolated AI agents — and why software sandboxes need more than…...
Key takeaways After Alibaba’s Qwen model tried a test upload to a server in Beijing during a Cohere experiment, we replayed that moment 64 times, and the model picked 52 different addresses. That points to an unpredictable model rather than a hidden ...
I wanted to let a coding agent change an endpoint while keeping application configuration out of reach. The request was fairly simple: update the Java source, adjust the tests, and leave the rest of the project alone. Writing that in the prompt expla...
“Vibe coding” has become an insult this month. The data points at something narrower and more fixable: six boundaries that AI-written apps…...
Key takeaways Only 24 percent of organizations can fully inspect AI traffic without impacting performance, according to Check Point’s 2026 Cloud Security Report. →Voice agents are easy to talk into things. Researchers got a leading realtime voice mod...
The hardware underpinning major AI models can be hacked so that chatbots provide faulty and dangerous answers to questions from users, according to new research shared exclusively with CNN....
New threat research on an attack chain targeting AI compute infrastructure, enabling compromise of a model with no visible IOCs....
Confidence scoring and human-in-the-loop routing let production AI flag uncertain outputs, route them to the right expert, and learn from every correction.
Bowen Wan’s parents allege ChatGPT guided her off seizure medication and contributed to her death. Their lawsuit puts AI medical advice…...
Authors: Alan Chan, Christoph Winter, Andrew Barto, Jakub Pachocki, Geoffrey Hinton, Eric Horvitz, Yoshua Bengio, Dawn Song, Jack Clark, Hilary Greaves, Anton Korinek, Samuel Hammond, Thore Graepel, Ben Bariach, Philip H. S. Torr, Sheila A. McIlraith...
Your AI attack surface has two parts: the AI itself, and the places where AI runs. Apps, models, agents, and Model Context Protocol (MCP) servers now sit on laptops and servers, in network traffic, and inside SaaS and cloud platforms. That makes them...
Anthropic’s S-1 puts a training pause, AI models reaching real systems, and safety-driven delays in front of investors. That moves AI…...
SecurityOct 29, 2026
Does ChatGPT have a watermark? Understand the EU rollout, API defaults and detector access, plus practical checks for hidden characters and misleading AI scores.
SafetyOct 29, 2026