AI agents are considered one of the biggest productivity tools for technology companies. Every tech company, from Microsoft to Amazon and others, is talking about the potential of AI agents. Salesforce CEO Marc Benioff even said the company might change its name to Agentforce. Last December, a Salesforce employee told Business Insider that the company's official name may change to Agentforce. In a subsequent interview, Salesforce CEO Benioff acknowledged that this was a possibility. “Maybe,” Benioff said, adding, “That doesn't shock me.” Salesforce's CEO says customers now don't want to hear about the cloud anymore, something he learned during a recent focus group ahead of the company's annual Dreamforce conference in October. Instead, clients want to talk about AI agents.“I completely deleted the word cloud,” he said. “We did this at Dreamforce. Did you notice we never used it in the keynote? We found in focus groups that customers don't talk about the cloud anymore. They just want to talk about the agent interface.” But Wendy Whitmore, chief security Intel officer at Palo Alto Networks, doesn't seem to agree with Benioff. In an interview with the Register, Whitmore said AI agents will become a new insider threat to businesses in 2026. “CISOs and security teams are under tremendous pressure to deploy new technology as quickly as possible, which puts teams under tremendous pressure and enormous workloads to quickly move through procurement processes and security checks to understand whether new AI applications are secure enough for these organizations' use cases,” Whitmore told The Register. “And that gave rise to the concept that AI agents themselves become a new insider threat,” she added.“When you look at it through a defender's lens, a lot of what agent capabilities allow you to do is start thinking more strategically about how you defend your network, instead of always getting caught up in a reactive situation,” Whitmore said.
When a Chinese spy “instructed” Claude to infiltrate about 30 companies.
Whitmore, like nearly every other cyber executive the Register spoke to for the report, cited the October 2025 “human attack” as an example. Chinese cyber spies used Anthropic's Claude Code AI tool to attempt to digitally infiltrate about 30 high-profile companies and government agencies, and government-sponsored snooping was “successful in a small number of cases,” according to a report from the AI company. The mid-September operation targeted large tech companies, financial institutions, chemical manufacturers, and government agencies.Whitmore doesn't expect AI agents to carry out fully autonomous attacks this year, but he does expect AI to become a power multiplier for network intruders. “You'll find that these really small teams are almost as capable as large armies,” she says. “They can now leverage AI capabilities to perform more work that previously would have required much larger teams to perform.”
