Everything about AI
Disk Space
Joined February 2023
- Tweets631
- Following271
- Followers78
- Likes535
“Can models take actions that differ from their stated reasoning?”
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.
Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.
Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.
But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?
We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.
Read the full report: cohere.com/blog/concerning-b…
AI : Quest retweeted
SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios.
The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios.
Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said.
Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior.
Read my latest for Axios here: axios.com/2026/10/22/ai-agen…
AI : Quest retweeted
DeepSeek & Huawei are teaming up on new AI chip designs to challenge Nvidia, per The Information.
DeepSeek would be working with China’s biggest chipmaker to reduce reliance on US chips.
It comes a week after China’s CXMT chip progress sparked a global sell-off in stocks.
Embracing Open-source 🫡
🚀 Introducing DeepInfer, our in-house inference framework, now fully open source!
Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable.
🔗 github.com/deepseek-ai/DeepI…
1/4
Watch out with AI related Github repo's 👇 bleepingcomputer.com/news/se…
OpenAI textGrain watermarking is rolling out for EU ChatGPT and Codex.
We're expanding our approach to content provenance to include text in response to EU regulatory requirements, while recognizing the significant limitations of current text watermarking technology.
Our tools already help verify whether an image or audio file was created with our models. This work builds on those efforts to help people better understand when content may have been generated or edited with an OpenAI model.
In the EU, we’ll start watermarking eligible text from ChatGPT and Codex over the coming weeks to comply with the EU AI Act.
Customers using our API can turn on text watermarking for select models worldwide today.
AI : Quest retweeted
🚨SHOCKING: ANOTHER OpenAI safety lead just QUIT, saying the company's "culture is broken."
David Robinson, who led OpenAI's launch safety reports, is the 4th safety staffer out in a week, after three were fired for allegedly leaking confidential info.
At least a dozen senior OpenAI leaders had already left this year by August, per CNBC.
He says he "never encountered a colleague who had experience making airplanes fly safely or nuclear reactors run without melting down."
There are however some privacy concerns when using AI (non-local) euobserver.com/239963/ai-cha…
AI : Quest retweeted
Hermes Agent has had some MASSIVE updates the last month NOBODY is talking about
- Bot mode
- Local AI mode
- Improved subagents
- Hermes HUD
and much more
It's the most powerful open source agent ever
In this video I cover EVERY new change and how to get the most out of them
🤓 Math never was my thing anyway.
JUST IN: NVIDIA $NVDA CEO Jensen Huang says we're entering a "low IQ era" where children won't need to learn basic math.
Check out @AnthropicAI 's threat intell report: anthropic.com/threat-intelli…
AI : Quest retweeted
🚨 BREAKING: US Treasury Secretary Scott Bessent declares OpenAI management personally responsible for HuggingFace hack
> "The Hugging Face incident, that is the responsibility of the OpenAI management, not a bunch of agents."
> "It is humans who are responsible, not the AI."
> "What we shouldn’t do on safety is to give these labs a liability exemption, which is what they are asking for."
> "The best way to guarantee safety is that the creators are liable for what they build and generate."
It’s OVER
Altman to brief the UN Security Council,
OpenAI’s CEO is scheduled to present next week on AI and global security. m.economictimes.com/tech/art…
AI : Quest retweeted
Palantir CEO Alex Karp on CNBC today: closed models can absorb a company’s proprietary knowledge into someone else’s system, while open-weight models let the company keep that knowledge, customization, and control inside its own environment.
"When you use a closed model and you pay for tokens, the reason the tokens are discounted is not so they can grow the market. They’re discounted so that they can get access to their IP, so that their models get better.
And there was functionally no honest conversation about this or disclosure.
Now, that’s point one.
Point two is: without an application layer and without post-training, which is currently being done with open-weight models on the classified side—and it’s just taking off across America because it’s cheaper and ends up being safer, and you control your future and your present"
----
From "CNBC Television" YouTube channel, (full video link in comment)