Source: Fortune
Summary
Top AI labs have called for slowing development to address safety concerns as AI agents act unpredictably. CIOs are integrating AI into operations while managing risks. Joe Atkinson of PwC warns enterprises must plan for AI risks. Cisco has centralized AI agents in its MyAgent platform, with 90,000 employees using it. Intuit built security into its AI system from the start. Workday and ServiceNow offer AI governance tools. Zscaler is part of an alliance to develop safe AI tools. Security experts say AI introduces new, unpredictable risks.
Our Reading
The numbers tell one story.
AI labs want to slow down.
CIOs are racing to control agents.
Cisco centralizes AI with MyAgent.
Zscaler joins an alliance for safer AI.
AI agents are going rogue, but companies are building guardrails.
CIOs are under pressure to manage risks.
Cisco limits third-party AI to avoid sprawl.
Intuit embeds security in its AI system.
Workday and ServiceNow offer governance tools.
Author: Evan Null
AI Labs Call for Caution
Some of the world’s top artificial intelligence labs have called for a slowdown in development, citing the need for more safety controls as AI agents act unpredictably. This comes as CIOs work to integrate AI into their operations while managing the growing risks of autonomous systems exploring vulnerabilities and outmaneuvering human monitoring.
CIOs Face Rising Risks
Joe Atkinson, global chief AI officer at PwC, says enterprises need to focus on AI risks and start planning for them. As AI agents proliferate, C-suite executives must collaborate to enforce guardrails, monitor AI activity, and track tasks. Department heads across the business must also play a role in managing digital employees.
Cisco Centralizes AI Control
Cisco has built its own AI agent platform, MyAgent, to centralize all company-authorized AI models and data. The platform is used by 90,000 employees, with 50% daily adoption within two weeks. Cisco’s executive vice president, Thimaya Subaiya, says the company will not authorize third-party AI agents to avoid “agent sprawl.”
Intuit Embeds Security From the Start
Intuit’s CTO, Alex Balazs, says the company built security, risk, and fraud controls into its AI system from the beginning. He warns that retrofitting security after the fact is not an option. Balazs also notes that AI agents going rogue have mostly occurred during testing, but he remains cautious about relying on AI to do the right thing.
Security Experts Warn of New Risks
Sam Curry, CISO at Zscaler, says AI introduces new, unpredictable risks. He calls AI a “new form of insider” that is non-deterministic and can take initiative. Zscaler has joined the Open Secure AI Alliance to develop safe AI tools. Curry warns that the “characteristic psychology of AI” is still not fully understood, and its incentives remain unclear.
