Nvidia releases tools to rein in rogue AI agents

AI nannies for AI agents?

"Nvidia released the Open Agent Safety Platform, software meant to help developers test and monitor AI agents and stop them from acting outside their intended boundaries.
Nvidia says the platform includes
(i) a watchdog that can quarantine misbehaving agents within milliseconds and
(ii) tools to control what agents can see, access, and interact with.

The release follows several incidents in which AI agents from OpenAI, Anthropic, Meta, and Google went rogue, including an OpenAI agent that Australian Prime Minister Anthony Albanese said infiltrated a government-services website, and Anthropic’s disclosure that its test software hacked outside companies without its knowledge in three cases since April.
Nvidia said each incident followed the same pattern: agents escaped sandbox safety controls to complete their assigned tasks. For developers building AI agents, the platform offers a vendor-backed option for enforcing boundaries and monitoring behavior in production; however, there is currently no independent verification of how well it prevents the sandbox-escape pattern Nvidia describes." (Data Points)

NVIDIA Open Agent Safety Platform: A Reference for Continuous In-Silicon Agent Monitoring "Secure AI agents with a safety enforcement layer spanning software and hardware"

Chipmaker Nvidia is spending $6 billion on AI

AI is not cheap! AI runs on powerful chips, but is a chipmaker a good candidate for advanced AI? Or is the charismatic CEO Jensen Huang going too far?

"That’s how much Nvidia is spending as it tries to build one of the world’s most powerful open-weight AI models. The effort is aimed at competing with Chinese heavyweights and American AI giants."

The Wall Street Journal What's news

Jensen Huang (official photo, source)

Israel’s GDP growth reflects fast-rising Nvidia revenue

Amazing stuff! Like Nvidia Manna from heaven! 😊

"When Nvidia acquired Mellanox six years ago, few officials at the Ministry of Finance and Bank of Israel would have imagined that within a short time, the US chipmaker would become not only a global phenomenon and the most valuable company in the world, but also a major engine behind the growth of the Israeli economy - all without actually producing a single chip in Israel itself.

With 6,000 employees in Israel, Nvidia's footprint in the economy has become so large that the Central Bureau of Statistics has produced a new figure for growth without Nvidia’s activity. ...

Over the entire previous year, Israeli GDP grew by 3.5%, but without Nvidia, growth would have amounted to only 2.1%. ..."

Israel’s GDP growth reflects fast-rising Nvidia revenue - Globes "Nvidia’s networking division is seeing an exponential rise in revenue, making it a major engine of growth for the Israeli economy."

How Open Models Are Driving AI Research: Nvidia as An example

Impressive! This is only the beginning of an exciting revolution in science and medicine!

"Every year, the International Conference on Machine Learning (ICML) reveals where thousands of AI researchers have decided to put their work. 

This year’s accepted papers reveal a clear direction: open frontier models and open AI infrastructure have become foundational to how modern AI science gets done.

NVIDIA had 74 papers accepted at ICML 2026. Approximately 2,000 accepted papers cite NVIDIA GPUs, and 145 cite NVIDIA Nemotron — a family of open models, including open datasets — as the foundation for new research.

Hundreds more draw on NVIDIA Cosmos, NVIDIA Isaac GR00T, BioNeMo and other NVIDIA open model families, spanning physical AI, robotics, autonomous vehicles and biomedical research.

The Themes Defining This Year’s Research

Areas including vision and video generation, reinforcement learning for large language models (LLMs) and agent training as well as AI inference remained prominent themes across this year’s papers, reflecting sustained investment these fields command — while several new areas also broke through.

Robot world models drew significant attention, with papers like DreamDojo pushing the boundary of how AI systems learn to reason about and act in physical environments. DreamDojo, for example, learns how the physical world behaves from human video and builds on NVIDIA Cosmos open frontier models to predict how a robot would handle objects and operate in environments it was never trained on. It lets researchers evaluate policies, plan actions and teleoperate a virtual robot, accelerating development without the costs and risks of physical deployment.

AI for life sciences was fueled by NVIDIA BioNeMo open models and research contributions that help researchers understand protein function, molecular behavior and genetic code. Papers like FLIP2 introduce public benchmarks for testing how well AI predicts the effects of protein mutations. KERMT is a new BioNeMo open model for predicting molecular properties important to drug discovery. 

Synthetic data generation (SDG) drew particular interest at ICML this year with several Nemotron and physical AI open datasets, reflecting a broader shift in how researchers are thinking about training at scale without relying solely on human-labeled data. ..."

How Open Models Are Driving AI Research | NVIDIA Blog "At the International Conference on Machine Learning, nearly 145 papers cite NVIDIA Nemotron open models and datasets as the foundation for new research."

International Conference on Machine Learning (ICML) 2026 (South Korea 7/6-7/11)

The great wipeout: Nvidia lost $1 trillion within two months

Wow! Roller coaster!

The great wipeout: Nvidia lost $1 trillion within two months | The Jerusalem Post "Nvidia recorded a 16% drop from its peak to its lowest level since 2019. But still, do not count it out, because its business performance is at a peak. So where did the investors' money disappear?"

Nvidia spending up to $150bn a year on Taiwan AI suppliers: Jensen Huang

A stunning headline!

"Nvidia CEO Jensen Huang declared Taiwan the heart of the AI revolution. He believes the island will be a global technology manufacturing center for years. Huang announced plans for Nvidia's new Taiwan headquarters. Groundbreaking is set for this year. The facility is expected to be operational by 2030. This marks a significant future investment. ...

""Four ⁠years ago, ⁠five years ago, Nvidia was spending about $10, $15 billion dollars a ​year in Taiwan. Now we're spending $100, going to $150 billion dollars in ​Taiwan each year," ..."

Nvidia spending up to $150bn a year on Taiwan AI suppliers: Jensen Huang - Nikkei Asia "CEO says company will quadruple hiring in AI 'epicenter' to 4,000"

Nvidia to spend $150 billion a year in Taiwan, 'epicentre' of AI revolution, says CEO

On Gated DeltaNet-2: Decoupling Erase and Write in Linear Attention

This could be an interesting paper by Nvidia researchers Yejin Choi and Jan Kautz!

"Linear attention replaces the unbounded cache of softmax attention with a fixed-size recurrent state, reducing sequence mixing to linear time and decoding to constant memory.

The hard part is not just what to forget, but how to edit this compressed memory without scrambling existing associations.
Delta-rule models subtract the current read before writing a new value, and Kimi Delta Attention (KDA) sharpens forgetting with channel-wise decay. But the active edit still uses a single scalar gate to control two different things: how much old content to erase on the key side and how much new content to commit on the value side.

We introduce Gated DeltaNet-2, which generalizes both Gated DeltaNet and KDA by inheriting adaptive forgetting and channel-wise decay while addressing their shared limitation, the scalar tie between erasing and writing.

Gated Delta Rule-2 separates these roles with a channel-wise erase gate b_t and a channel-wise write gate w_t, reducing to KDA when both gates collapse to the same scalar and to Gated DeltaNet when the decay also collapses.

We derive a fast-weight update view, a chunkwise WY algorithm with channel-wise decay absorbed into asymmetric erase factors, and a gate-aware backward pass that preserves efficient parallel training.

At 1.3B parameters trained on 100B FineWeb-Edu tokens, Gated DeltaNet-2 achieves the strongest overall results among Mamba-2, Gated DeltaNet, KDA, and Mamba-3 variants across language modeling, commonsense reasoning, and retrieval.

Its advantage is most pronounced on long-context RULER needle-in-a-haystack benchmarks, where it improves the evaluated multi-key retrieval setting and remains strong in both recurrent and hybrid settings. ..."

Data Points

[2605.22791] Gated DeltaNet-2: Decoupling Erase and Write in Linear Attention

Chart of the day

The phenomenal revenue growth of Nvidia! "Sales for the April quarter reached $81.6 billion, up 85% from the year-earlier period " (Source)

Chart generated by Google AI 

Nvidia’s Jensen Huang joins Trump’s trip to China

Good news! This promises to be a very interesting summit! Expectations are rising!

"... Seventeen American CEOs were on the original list of business delegates released by the White House on Monday – a smaller group than Trump’s previous China trip in 2017, when 27 high-profile executives joined. ..."

Breaking | Nvidia’s Jensen Huang joins Trump’s trip to China after all | South China Morning Post "The Nvidia CEO, not included in the White House’s original delegation list, will be in Beijing with the US president, the company has confirmed"

Notes on ProRL Agent: Rollout-as-a-Service for RL Training of Multi-Turn LLM Agents

This could be an interesting paper by Jan Kautz of Nvidia and co-authors!

"Multi-turn LLM agents are increasingly important for solving complex, interactive tasks, and reinforcement learning (RL) is a key ingredient for improving their long-horizon behavior. However, RL training requires generating large numbers of sandboxed rollout trajectories, and existing infrastructures often couple rollout orchestration with the training loop, making systems hard to migrate and maintain. Under the rollout-as-a-service philosophy, we present ProRL Agent , a scalable infrastructure that serves the full agentic rollout lifecycle through an API service. ProRL Agent also provides standardized and extensible sandbox environments that support diverse agentic tasks in rootless HPC settings. We validate ProRL Agent through RL training on software engineering, math, STEM, and coding tasks. ProRL Agent is open-sourced and integrated as part of NVIDIA NeMo Gym."

[2603.18815] ProRL Agent: Rollout-as-a-Service for RL Training of Multi-Turn LLM Agents

'HUGE BREAKTHROUGHS': NVIDIA CEO Jensen Huang reveals future of AI with Maria Bartiromo

Recommendable! If only half comes true what he has mentioned in this interview then it would still be an economic boom!

Nvidia launches Alpamayo, open AI models that allow autonomous vehicles to 'think/judge like a human'

Good news!

Perhaps, I do not have to renew my driver's license anymore! Just kidding!

I wish it was not thinking/judging like a human as I do not like to get a ticket from or be pulled over by a (robot) police officer! 😊

"At CES 2026, Nvidia launched Alpamayo, a new family of open source AI models, simulation tools, and datasets for training physical robots and vehicles that are designed to help autonomous vehicles reason through complex driving situations. ..."

"News Summary:

• NVIDIA is the first to release an open reasoning VLA model designed to tackle long-tail autonomous driving challenges; NVIDIA Alpamayo family also includes simulation tools and datasets for AV development.
• Alpamayo 1, AlpaSim and Physical AI Open Datasets enable the development of vehicles that perceive, reason and act with humanlike judgment [???]  — enabling developers to fine-tune, distill and test models that unlock greater safety, robustness and scalability.
• With Alpamayo, mobility leaders such as JLR, Lucid and Uber, along with the AV research community including Berkeley DeepDrive, can fast-track safe, reasoning‑based level 4 deployment roadmaps.

..."

Nvidia launches Alpamayo, open AI models that allow autonomous vehicles to 'think like a human' | TechCrunch

NVIDIA Announces Alpamayo Family of Open-Source AI Models and Tools to Accelerate Safe, Reasoning-Based Autonomous Vehicle Development (original news release)

Nvidia buying AI chip startup Groq for about $20 billion, biggest deal ever for Nvidia

Good or bad news! Just before Christmas so it will not get noticed as much? This is huge!

Was it a straight direct buy or a license agreement plus poaching the top employees?

"Key Points

• Nvidia is making its largest purchase ever, acquiring assets from nine-year-old chip startup Groq for about $20 billion.
• The company was founded by creators of Google’s tensor processing unit, or TPU, which competes with Nvidia for artificial intelligence workloads.
• Groq, which was valued at $6.9 billion in a financing round in September, framed the deal as a “non-exclusive licensing agreement,” with its CEO and other senior leaders joining Nvidia.

..."

Nvidia buying AI chip startup Groq for about $20 billion, biggest deal

Credits: Last Week in AI

Source

National Review: Trump’s China Chip Deal & Sale of American National Security. Really!

This article is the editor's pick for today of the National Review!

Well, the chips (H200 AI chip) now allowed to be sold to China are not the latest and best chips ((H100/H200, B200 GPU)!

Trump’s China Chip Deal & Sale of American National Security | National Review "If letting China buy precious tech is detrimental to U.S. interests, that threat is not mitigated by allowing the Treasury to grab a quarter of the income."

How Nvidia GPUs Compare To Google’s And Amazon’s AI Chips and other chips

Recommendable!

Why West Asia Is the New Front in the AI Chip War with Palki Sharma

Very recommendable!

Nvidia, Deutsche Telekom strike €1B partnership for a data center in Munich

Is Germany a developing country when it comes to machine learning & AI? Possibly!

"The company on Tuesday signed a €1 billion ($1.15 billion) partnership with Deutsche Telekom to set up an “AI factory” in Munich that aims to boost Germany’s AI computing power by 50%.

Called the “Industrial AI Cloud,” the project will use more than 1,000 Nvidia DGX B200 systems and RTX Pro Servers with up to 10,000 Blackwell GPUs to provide AI inferencing and other services to German companies while complying with German data sovereignty laws. ..."

Nvidia, Deutsche Telekom strike €1B partnership for a data center in Munich | TechCrunch

Microsoft's $15.2B UAE investment turns Gulf State into test case for US AI diplomacy

Good news! How much was President Trump behind this?

Maybe this will also further promote the economic cooperation between Israel and Arab countries!

"Microsoft will invest $15.2 billion in the United Arab Emirates over the next four years, the company announced Monday at the first annual Abu Dhabi Global AI Summit. The investment will include the first-ever shipments of the most advanced Nvidia GPUs to the UAE.

As part of the deal, the U.S. has granted Microsoft a license to export Nvidia chips to the UAE, a move that positions the country as both a proving ground for U.S. export-control diplomacy and a regional anchor of American AI influence. ..."

Microsoft's $15.2B UAE investment turns Gulf State into test case for US AI diplomacy | TechCrunch

Nvidia Just Bet $1 Billion on Nokia. Here’s Why. with Palki Sharma

Good news! What an amazing comeback by Nvidia! Finnish, but not finished!

The $1 Trillion Tangled Web Of AI Deals Mapped Out

Recommendable! At the core are Nvidia and OpenAI.