Cybersecurity stocks were the standout winners on Wall Street on Monday, September 28, as investors bet that the rapid spread of autonomous AI agents will create a much bigger market for digital security tools.
Palo Alto Networks (PANW), CrowdStrike (CRWD) and Zscaler (ZS) finished as the top three performers in the Nasdaq-100, gaining 4.6%, 2.8% and 3.3% respectively, even as the broader technology sector came under pressure.
All three shares are now trading near all-time highs, extending a broader rally in the sector. The move came after Nvidia announced its Open Agent Safety Platform, a new open-source framework designed to provide security controls around autonomous AI agents from testing through deployment.
The shares dipped 0.1% to 0.4% in overnight trading on Tuesday, but held most of their gains.

Nvidia's Open Safety Push Sets the Tone

Nvidia said CrowdStrike and Palo Alto Networks are among more than 100 organizations working with it on the platform, which aims to create a shared trust layer for safe agent systems.
Announcing the initiative, Nvidia chief executive Jensen Huang wrote in a post on X: "This is bigger than a single product. It's the beginning of an open ecosystem to build the trust layer for safe agent systems."
Analysts said the framework validated what cybersecurity investors have been arguing for months: that agentic AI is not just another software trend, but a structural expansion of the attack surface — and of the market for defending it.
As agents move from simple chatbots to systems that execute multi-step workflows across databases and production tools, every new deployment becomes a customer for monitoring, policy enforcement and runtime protection.

Safety Scares Sharpen the Bid

A string of recent AI safety incidents has sharpened the bid. OpenAI disclosed in July that an AI agent being tested internally escaped its containment and compromised the infrastructure of AI platform Hugging Face during a cybersecurity evaluation.
More recently, OpenAI agents attempted unauthorized access to government and institutional websites, including U.S. government platforms and an Australian government portal. Anthropic has disclosed multiple incidents in which its Claude models gained unauthorized access to real third-party systems during cybersecurity evaluations.
On Monday, OpenAI also canceled the planned October release of its GPT-6.1 Astra model after internal testing raised concerns around safety and alignment, including instances in which the model could evade human oversight.

A Broader Rotation Into Cyber

The session also reflected a broader rotation within technology, with investors trimming profits from hardware and chipmaker giants to fund positions in cloud and network defense vendors.
Palo Alto Networks and CrowdStrike have extended a broader multi-month rally, with both sitting near recent highs as Wall Street reframes AI spending: the bet is no longer only on who builds the smartest models, but on who keeps them under control.
By Hamza Faisal