AI Has to Be Secured by Design Says Palo Alto CEO as Company Acquires Protect AI

Palo Alto Networks is acquiring Protect AI, a fast-rising AI security startup based in Seattle, in a deal valued at over $500 million. First reported by GeekWire, the transaction is expected to close later this year, pending customary regulatory approvals. Once complete, Protect AI’s 120-person team including its founding leadership will join Palo Alto Networks’ expanding security portfolio.

Protect AI was founded in 2022 by Ian Swanson, Badar Ahmed, and Daryan Dehghanpisheh, former engineering leaders from Amazon, AWS, and Oracle. The company builds tools that monitor and secure every layer of the AI/ML development lifecycle, from data ingestion to deployment. Its software focuses on securing machine learning systems from risks such as training data poisoning, prompt injection, model manipulation, and AI supply chain vulnerabilities—areas where traditional security tools offer limited coverage.

The acquisition is directly tied to Palo Alto’s launch of Prisma AIRS, short for AI Runtime Security, which debuted at the RSA Conference in San Francisco. The new platform offers protection across the AI lifecycle and includes features like model scanning, runtime threat detection, AI red teaming, posture management, and autonomous agent security.

“As AI-powered applications become core to businesses, they bring risks traditional security tools can’t adequately handle,” said Anand Oswal, SVP and GM at Palo Alto Networks. “By extending our AI security capabilities to include Protect AI’s innovative solutions, businesses will be able to build AI applications with comprehensive security.”

Protect AI had raised a $60 million Series B round in 2024, bringing its valuation to approximately $400 million, according to Bloomberg. Investors include Salesforce Ventures, Samsung’s Venture Investment Corp., and 01 Advisors, a VC firm co-founded by former Twitter executives. Over the past year, Protect AI also acquired Laiyer AI, SydeLabs, Rebuff, and Huntr to strengthen its capabilities in red-teaming large language models and protecting open-source AI pipelines.

“Our mission has always been to help organizations secure the most important technology of our time: AI,” said Ian Swanson, Protect AI’s co-founder and CEO, in a LinkedIn post. On the acquisition, Swanson noted: “Joining forces with Palo Alto Networks will enable us to scale our mission… and amplify our impact globally.”

For Palo Alto Networks, the acquisition is part of a deliberate move into a rapidly growing segment of cybersecurity. CEO Nikesh Arora framed the timing and fit in an interview with CNBC.

“You can’t walk around any technology street corner without hearing the words ‘AI’ in your ear,” Arora said. “Part of our ethos has always been—when you deploy technology, you have to deploy it in a secure fashion. So we’re saying AI is secured by design.”

He added, “Our job is to get ahead of the technology trend and deliver a cybersecurity platform that allows our customers to, as we say, deploy bravely. After a rigorous analysis around the market, we discovered Protect AI was doing something complementary to us and actually fitting perfectly in our plans and the way we think about AI.”

Security researchers have warned that AI adoption is expanding faster than enterprises are prepared to secure it. A 2024 report from HiddenLayer found that nearly 75% of companies experienced at least one AI-related breach in the past year. Morgan Stanley, in its latest forecast, projects that the AI cybersecurity market will reach $135 billion by 2030.

“We’re in the early days,” Arora said. “I think we went through the cycle where people got really excited about using things like ChatGPT or Gemini or Claude—whichever model that has come up with a commercial user interface. We’re seeing people do amazing use cases—like make videos on the fly, transform pictures. All of that is going to start getting into some serious applications on the enterprise side. Enterprises are out there evaluating how to deploy AI, understand AI, make it work. Obviously, with all these new technological capabilities come challenges from a security perspective.”

He listed the key concerns: “How do you make sure the model doesn’t go rogue on you? How do you make sure it doesn’t get poisoned by training data? How do you make sure a model or an AI bot doesn’t get hijacked by bad actors?”

Arora also addressed the shift toward open-source models and the security gaps they introduce. “It’s a wild west,” he said. “There are tons and tons of models out there… Every one of those models needs to be scanned. You have to understand: Are there any vulnerabilities in those models? Who do they talk to in the backend? Is any training data getting leaked?”

He explained that Palo Alto’s AI firewalls are designed to help customers use both open- and closed-source models more securely. Synthetic data introduces another layer of complexity.

“Most models are training on widely available data,” Arora noted. “But there’s a large trove of proprietary data—data that belongs to corporations and enterprises—which will now start getting cleansed and put together to make it very useful.” He cited drug discovery and cybersecurity as examples of domains where private datasets, when combined with public information, can produce more specific and impactful outcomes.

The company is also watching closely as government cybersecurity budgets evolve. Cybersecurity job openings in the federal sector dropped 45% year-over-year, and spending fell 38% in Q1. Arora said the cutbacks are affecting people-centric services more than product investments.

“The federal government has spent the majority of its cyber dollars on services and people—and very little on product,” he said. “The only way to deliver the same quality and capability of cybersecurity that we need to protect our country… is to increase the product component.”

The Protect AI deal is the latest in a line of strategic acquisitions for Palo Alto Networks. The company has completed 21 deals to date, most recently acquiring Talon Cyber Security for $600 million in late 2023. Other acquisitions include Dig Security, Cider Security, Zycada Networks, and Bridgecrew.

With Protect AI, Palo Alto gains not just a product stack but a globally distributed team based in Seattle, Berlin, and Bangalore.“We’re delighted to embrace about 100 people from Protect AI, their customers, and all the innovations they have,” Arora said, “and put it together in our platform.”