Weekly Roundup: Nous Research Hits $1.5 Billion, Google Ships Gemini 4 Argon, and Workday Makes Agent Governance the Default

This week in agentic AI, the gap between capability and control widened in both directions. On one side, Nous Research closed a $90 million Series B at a $1.5 billion valuation, proving that open-source agents are now a credible enterprise alternative to closed platforms. On the other, India’s computer emergency response team reported that rogue AI agents had accessed live organisational systems through exposed credentials, showing exactly what happens when agent deployment outpaces security discipline.

In between, Google released Gemini 4 Argon globally for long-horizon agent tasks, Workday made its Sana interface the governed front door for HR and finance agents, and DXC Technology partnered with LangChain to bring agent orchestration into mission-critical IT operations. The pattern is consistent with recent weeks: the platform layer is maturing, the governance layer is becoming a procurement requirement, and the security layer is catching up to real incidents rather than theoretical risks.

For buyers, the message is that agent adoption is no longer a question of whether the technology works. It is a question of whether your organisation can secure, govern and observe it at scale. For suppliers, the competitive bar is shifting from model performance to operational reliability, auditability and integration with existing enterprise controls.

1. Nous Research raises $90 million: open-source agents enter the enterprise mainstream

On 7 October 2026, Nous Research announced a $90 million Series B round at a $1.5 billion valuation, with strategic investment from NVIDIA and Microsoft’s M12 fund. The company plans to expand Hermes, its open-source AI agent, into an enterprise product called Hermes for Businesses, allowing organisations to deploy customised agents on their own infrastructure.

Nous Research reports that Hermes has been cloned more than 24 million times and drives about 2.5% of global AI agent traffic. The funding signals a structural shift: open-source agents are no longer a research curiosity or a hobbyist project. They are becoming a procurement option for enterprises that want vendor independence, data sovereignty and the ability to customise agent behaviour without platform lock-in.

Why it matters: for buyers, the open-source versus closed-platform decision is now a real trade-off, not a theoretical one. Procurement teams will need governance frameworks that work across both proprietary and open-source agents. For suppliers, Nous Research’s valuation puts pressure on closed-platform vendors to demonstrate why their walled garden delivers enough value to justify the premium and the dependency.

2. Google releases Gemini 4 Argon: a model built for long-running agent work

On 14 October 2026, Google released Gemini 4 Argon globally for developers and enterprises through the Gemini API, AI Studio and the Gemini platform. The model is explicitly positioned for complex, long-horizon workflows such as code migration, enterprise research and cybersecurity analysis. Pricing starts at $2 per million input tokens and $10 per million output tokens, making small pilots relatively inexpensive while still requiring cost monitoring at scale.

Google also added a task mode to the Gemini app, allowing the assistant to continue working when the user is away, and connected it to more than 40 applications including Instacart, OpenTable, Adobe and Shopify. A separate announcement outlined a universal Gemini agent for startup teams, with cryptographic identity, isolated sandboxing and an Agent Gateway for governance, though these capabilities were not yet publicly available at launch.

Why it matters: buyers can now test agent workflows against a model designed for multi-step autonomy rather than simple chat. The risk is that longer autonomous runs increase total task cost, tool-call volume and review overhead. Buyers should measure end-to-end workflow cost and failure recovery, not just token pricing. Suppliers building on the Gemini platform should prepare for questions about sandboxing, identity and audit trails that Google has promised but not yet shipped.

3. Workday turns Sana into a governed front door for enterprise agents

Workday made Sana its primary agentic interface for HR, finance and IT this week, allowing employees to ask for outcomes in natural language while Sana coordinates the appropriate agents, tools and people. The company also launched its Agent Harness generally, with memory, guardrails, regional data controls, model choice, evaluations and step-by-step traces for auditing.

Importantly, Sana can operate across Workday and tools employees already use, including Claude, Microsoft 365 Copilot, Teams, Slack and Google Gemini Enterprise, while preserving permissions and approvals. This means Workday is positioning itself less as a collection of assistant features and more as an operating layer for controlled business-process automation.

Why it matters: enterprise buyers can evaluate Workday as a governed automation platform rather than a chatbot upgrade. The audit trail and permission model are particularly relevant for payroll, recruiting, employee data and financial close workflows, where mistakes are consequential and regulators are watching. For suppliers, Workday’s move sets a benchmark: if your agent platform cannot demonstrate equivalent governance, expect longer sales cycles and higher security scrutiny.

4. CERT-In reports rogue AI agents accessing live enterprise systems

On 16 October 2026, India’s Computer Emergency Response Team (CERT-In) said it is investigating incidents in which AI agents reached live organisational systems and accessed business records, including personal and confidential data. The reported root causes were not exotic: exposed credentials and keys in source code, repositories, configuration files, client applications and public web content.

The incidents demonstrate that agentic systems increase the speed and scale at which familiar security mistakes can be exploited. An exposed API key that might have sat dormant for months can now be discovered and used by an autonomous agent in minutes. CERT-In’s advisory recommends immediate scanning of repositories, configuration files, cloud roles and agent tool definitions for exposed secrets, followed by rotation and log review for unusual automated access.

Why it matters: this is the kind of incident that moves agent security from a procurement checkbox to a board-level concern. Buyers should treat agent tools and development environments with the same security rigour as internet-facing applications, rather than assuming that experimental agents are isolated. Suppliers should expect customers to ask for proof of secret management, runtime sandboxing and incident response capabilities before purchase.

5. DXC and LangChain package agent development for mission-critical IT operations

DXC Technology and LangChain announced a partnership this week combining DXC’s OASIS IT-operations platform with LangGraph for agent orchestration and LangSmith for development, evaluation and monitoring. OASIS is scaling toward 125 customer environments across regulated industries, with agents investigating operational events and bringing humans into decisions when judgement or approval is needed.

The partnership offers a practical pattern for deploying agents in environments where reliability and accountability matter more than an impressive demo. Teams get a framework for testing and observing agent behaviour, rather than treating production monitoring as an afterthought. DXC’s focus on regulated industries signals that even the most conservative sectors are now evaluating agentic AI for operational use.

Why it matters: buyers in regulated industries now have a reference architecture for agent deployment that includes evaluation, monitoring and human-in-the-loop governance. Suppliers should prepare for requirements that include replayable evaluations for common scenarios, explicit escalation paths and auditable decision trails. The partnership also shows that agent orchestration is becoming a procurement category in its own right, separate from both the model layer and the application layer.

The Agentic Expo takeaway

This was the week that enterprise agentic AI proved it can attract billion-dollar valuations, ship production-grade governance and trigger real security incidents, all at the same time. Nous Research’s funding, Google’s model release and Workday’s platform expansion show that the supplier market is maturing fast. CERT-In’s warning shows that the buyer risk is maturing just as quickly.

For buyers, the practical agenda is clear. Before deploying any agent at scale, verify that your secrets are rotated, your governance framework is enforceable and your monitoring stack can observe autonomous actions in real time. For suppliers, the message is equally clear: the best agents are the ones that work, explain themselves, survive failure and fit into the governance controls that enterprises already require.

The agentic economy is here. The question this week is not whether your organisation should participate. It is whether your operational layer is ready for the consequences.