Agents Go Autonomous. Control Planes Rise. Security Rewrites.

What happened
Palo Alto Networks announced plans to acquire Portkey, an AI‑Gateway startup whose platform serves as a central control plane for autonomous agents. Portkey’s gateway provides least‑privilege access enforcement, semantic routing with automated failovers and a 99.99 % uptime service, and caches agent artifacts to improve reliability and reduce costs.

Why it matters
As organizations move from copilots to fully autonomous agents, they need secure gateways that authenticate and govern agent actions. Palo Alto’s acquisition positions the cybersecurity giant at the heart of agentic infrastructure, signalling that AI gateways will be as critical as firewalls in the agentic era.

What’s next
The deal is expected to close later in 2026. Portkey will become the AI gateway within Palo Alto’s Prisma AIRS platform, giving enterprises an integrated path to deploy and manage autonomous agents securely. Expect further consolidation as security vendors race to own the control plane for AI workloads.

What happened
Okta published a “blueprint for the secure agentic enterprise” and introduced Okta for AI Agents, a platform that discovers and registers both sanctioned and shadow agents, standardizes agent access and provides a kill switch for rogue agents. The blueprint defines three questions—where are my agents, what can they connect to and what can they do—and offers capabilities like a universal directory for non‑human identities, an Agent Gateway to manage access tokens and a universal logout that can revoke all of an agent’s credentials.

Why it matters
Powerful “superagents” can now execute terminal commands, access file systems and spawn teams of sub‑agents, rendering human‑oriented identity models obsolete. Research cited by Okta shows 88% of organizations have experienced suspected AI agent security incidents, yet only 22% treat agents as independent identities. By giving every agent a first‑class identity and a governed lifecycle, Okta aims to make identity the new firewall for AI.

What’s next
Okta for AI Agents becomes generally available today and will integrate with partners like Boomi and DataRobot. As regulators tighten AI‑safety rules and enterprises grapple with “shadow agents,” other identity providers will likely follow with comparable frameworks.

What happened
Mirantis launched Lens Agents, an early‑access platform for running AI agents across desktop and cloud environments. The system connects desktop tools like Claude, Cursor, and Copilot, external autonomous agents and agents built within Lens, and applies enterprise‑grade identity and access controls, sandboxed execution, server‑side credential injection, full audit trails, policy‑based governance and active cost controls.

Why it matters
AI agents are proliferating outside centralized control; many run on laptops with access to multiple clouds. Lens Agents offers a unified control plane that governs any agent, any model and any environment under the customer’s policies. It extends Lens beyond its Kubernetes IDE roots into a broader platform for governed AI operations.

What’s next
Lens Agents supports compliance frameworks like SOC 2 and ISO 27001 and is aligned with emerging regulations such as the EU AI Act. Mirantis plans to expand the platform’s capabilities as more enterprises demand cross‑cloud agent governance.

What happened
Gen Digital, the parent company of Norton, Avast and LifeLock,unveiled VPN for Agents and expanded the Norton AI Agent Protection suite. The first consumer‑grade AI‑native VPN isolates agent traffic with multiple tunnels and an agent‑native design that doesn’t require downloads. Norton’s protection now monitors agent actions, blocks malicious prompts and scans code and files for threats.

Why it matters
Consumer AI agents are beginning to read email, manage finances, execute code and control smart‑home devices. Traditional VPNs can’t distinguish between human and agent traffic, leaving users exposed to prompt injections and exfiltration. By separating and securing agent connections and adding real‑time protection, Gen is building a trust layer for the consumer agentic ecosystem.

What’s next
VPN for Agents is available through the Gen Agent Trust Hub, and Norton’s protection will be baked into Norton 360. Expect consumer cybersecurity vendors to follow suit as autonomous agents handle increasingly sensitive tasks.

What happened
Accenture invested in Netomi and formed a strategic partnership to embed Netomi’s agentic AI platform into the consulting giant’s customer‑experience offerings. Netomi’s no‑code platform orchestrates multiple AI agents to anticipate needs, take actions and deliver personalized responses across chat, email and voice while maintaining governance and brand compliance.

Why it matters
As support volumes rise and customers expect instant resolution, agentic AI systems can free human agents for complex tasks and improve consistency. Accenture’s endorsement signals that orchestrated, multi‑agent platforms are becoming mainstream in enterprise customer experience. The partnership also means Accenture clients get direct access to Netomi’s tools and training.

What’s next
Netomi’s technology will be integrated into Accenture’s digital core, enabling end‑to‑end agentic service systems. Expect other consultancies and contact‑center vendors to seek similar partnerships or acquisitions as agentic customer service platforms compete for enterprise adoption.

What happened
A joint report by ICIMS and Aptitude Research found that 74% of job candidates use AI tools, yet only 18% of organizations use AI broadly in hiring. Talent leaders confused AI with automation, 58 % conflated the two and reported using AI mainly for candidate screening (58% and communication (54%). While 46% plan to adopt agentic AI for recruitment, 82% say governance is important but 45% lack formal frameworks.

Why it matters
The adoption gap suggests employers are lagging behind candidates in leveraging AI, risking slower hiring and poorer candidate experiences. Interest in agentic AI signals that recruiters want systems that can autonomously manage workflows, but confusion and governance gaps could derail progress.

What’s next
Companies may accelerate AI adoption in recruiting, with greater emphasis on integrated platforms and clear governance. Vendors are likely to offer agentic recruiting tools, and regulators may push for transparency around automated hiring decisions.

What happened
Norway‑based robotics company 1X opened a 58,000‑square‑foot factory in Hayward, California to produce its NEO humanoid robots. The facility integrates motor, battery and sensor manufacturing under one roof and can produce 10,000 robots annually, with plans to scale to more than 100,000 units by 2027. Early demand sold out the first year’s production, and NEOs run on NVIDIA Jetson Thor computing, using Isaac simulation tools for training.

Why it matters
This is the United States’ first vertically integrated humanoid‑robot factory, marking a shift from prototypes to mass production of physical AI. Vertical integration reduces supply‑chain dependencies and accelerates iteration, while the use of high‑performance AI hardware enables real‑time autonomy for household robots.

What’s next
1X will ramp up production and deliver NEOs to early customers, while competitors like Agility Robotics and Figure are also scaling up. Expect rapid improvements in capabilities and price as humanoid robots enter homes and workplaces.

What happened
Reuters reports that in central Ukraine, human deminers are now working alongside AI-powered machines and robots to accelerate the removal of landmines. The collaborative effort leverages advanced robotics and AI to improve safety and efficiency in hazardous environments

Why it matters
This deployment shows embodied AI’s growing real-world impact in high-risk, mission-critical scenarios—moving beyond factories into humanitarian operations. It’s a powerful example of man-machine teaming for public good.

What’s next
Expect further scaling of AI-robotics partnerships in conflict zones and disaster recovery, with lessons learned likely to inform future standards for safety and autonomy.

As agents move from assistants to autonomous actors, the real battle is shifting to control—identity, gateways, and governance become the new infrastructure. The winners won’t just build smarter agents; they’ll own the systems that decide what those agents are allowed to do.

Spin up a simple multi-agent workflow and then add control.

Start by using Netomi (or a no-code agent builder) to create a basic customer support agent that can answer questions and take a simple action (like drafting a response or routing a request). Then introduce governance: map out what that agent should and shouldn’t be allowed to access—files, APIs, or actions.

Next, layer in identity and control concepts inspired by Okta, treat your agent like a user. Define its “role,” restrict permissions, and create a manual “kill switch” (even if it’s just turning off API keys or access tokens).

Finally, simulate scale: add a second agent or workflow step and think through how they coordinate and where you’d need a gateway layer like Palo Alto Networks is building with AI gateways.

The goal isn’t perfection, it’s to feel the shift: building agents is easy; controlling them is the real work.