Law360 (October 19, 2026, 9:30 PM EDT) -- Cisco Systems, Fortinet and a company accusing them of patent infringement told a judge Monday that the two sides are "diametrically opposed" on whether source code at issue should be analyzed with artificial intelligence, with the defendants fearing that AI could "escape the sandbox," as in recent high-profile hacks.

The two tech companies and plaintiff Sectra Communications AB, a Swedish cybersecurity provider, said they met Friday and were unable to reach an agreement on Sectra's motion to compel the use of AI tools to analyze the defendants' source code. They said that they "could not achieve common ground or otherwise resolve their dispute."

"This is an important issue to both sides, and they are diametrically opposed," the parties told Eastern District of Texas Judge Robert Schroeder III.

The judge has scheduled a hearing on Sectra's motion for Nov. 10, as the parties dispute the relevance of a spate of headlines about rogue AI agents from OpenAI, Anthropic and Google accessing the internet without authorization and hacking other systems.

Cisco and Fortinet told the judge last month that Sectra's request to "use AI-based tools poses both known and unknowable risks to the security of defendants' source code." They called that code their "crown jewels," saying that the inadvertent disclosure of it "would cause enormous, irreparable damage."

Cisco and Fortinet expressed concern that the AI tools could put the source code at risk if they "escape the sandbox" and access the internet without authorization, as happened in the recent hacks.

"Plaintiff asks this court to nonetheless blaze a new trail by endorsing the safety of these AI-based tools," the companies said, arguing that "any benefit from these tools is far outweighed by the risks imposed on defendants' businesses."

Sectra, which alleges in separate suits that Cisco and Fortinet infringed its wireless communication patent, moved last month for the judge to require the companies to allow Sectra to use AI tools to analyze the source code for the allegedly infringing technology.

AI software like Cline, Ollama and Qwen are "quite simply, the latest and greatest review tools available," Sectra said, so analyzing "massive source code repositories without these tools is disproportionately burdensome."

As for the concern raised by Cisco and Fortinet about an "AI breakout" potentially exposing their source code, Sectra said "it is impossible" for that to happen in this case, because the computers being used wouldn't be connected to the internet.

The defendants, Sectra said, cite nothing more than a "generalized fear of AI," but the protective order in the case "does not permit defendants to withhold approval based on stigma. It requires a reasonable basis. Defendants have none."

Cisco and Fortinet contended that Sectra's motion "ignores the recurring headlines" about AI agents finding ways of "breaking free from restricted environments even though engineers believed there was no internet connection."

Sectra's assurances that the AI tools it wants to use won't be able to do that "are as cavalier as they are uncredentialed," Cisco and Fortinet said. "With AI, the human reviewer does not know exactly what the tool will do."

In a reply brief, Sectra maintained that in the recent high-profile cases, the AI agents did in fact have internet access, which was unintentionally made available to them, so "none involved a non-networked environment." There would be no risk of an AI breakout if the source code analysis tools were used on a computer with no internet access, Sectra said.

It argued that "defendants' position is incredible in light [of] defendants' own use of AI."

"Defendants publicly proclaim that they use AI to develop and test their own code and instruct their customers on how to use AI safely, but then tell the court that Sectra's tools are too risky because they are AI," the brief said. "This is unreasonable."

Cisco and Fortinet contended that their "extensive knowledge of both network security and the risks and rewards of AI is precisely why defendants' position on the review of their source code with AI is so firm."

Sectra sued Cisco and Fortinet in March, alleging that the companies sell network security products that infringe a patent Sectra acquired in 2019 when it bought a company called Columbitech AB.

According to the suits, Sectra previously licensed Columbitech's patent, which covers technology that supports the ability of mobile devices to switch from using Wi-Fi to cellular internet connections.

Sectra said its Tiger security communications system incorporates the patented technology and is used by clients including the Norwegian Armed Forces, Sweden's defense ministry and NATO.

On Thursday, Sectra filed a new suit against Verizon over the same patent. It previously sued Absolute Software Inc. over the patent, but the parties reached an agreement in 2024 to dismiss that suit with prejudice in the Western District of Washington.

The patent at issue is U.S. Patent No. 7,797,437.

Counsel for the parties could not immediately be reached for comment Monday.

Sectra is represented by Edward Nelson, Brent Bumgardner, Christopher Granaghan, Carder Brooks and Jinming Zhang of Nelson Bumgardner Conroy PC, and Andrea Fair of Miller Fair Henry PLLC.

Cisco and Fortinet are represented by Matthew Gaudet, L. Norwood Jameson, Alice Snedeker, Glenn Richeson, Holly Engelmann, Joseph Powers, Brianna Vinci and David Dotson of Duane Morris LLP, and Melissa Smith of Gillam & Smith LLP.

The cases are Sectra Communications AB v. Cisco Systems Inc., case number 2:26-cv-00174, and Sectra Communications AB v. Fortinet Inc., case number 2:26-cv-00175, in the U.S. District Court for the Eastern District of Texas.

--Editing by Melissa Treolo.

For a reprint of this article, please contact reprints@law360.com.