At the CyberRisk Collaborative's NYC Leadership Exchange, the former Air Force intelligence and cyber leader told security executives that AI-manufactured information is steering boardroom decisions, and few organizations know who is responsible for catching it.
NEW YORK — For most of her 34-year Air Force career in intelligence and cyber operations, retired Lt. Gen. Mary O'Brien's job was keeping data in. Speaking last week at the CyberRisk Collaborative (CRC) NYC Leadership Exchange, she said that mission has reversed.
"One of the most important things I think people should think about today is what I'm calling 'The Flip,'" O'Brien told an audience of CISOs and senior security leaders. Exfiltration is still a risk, she said, but the larger threat is bad external data, created at scale with AI, reaching executives who then act on it instead of on their own trusted internal analysis.
The gap, O'Brien argued, is built into the discipline itself. "We've built zero trust for identities, devices and code. We still extend implicit trust to information," she said. "Integrity means not altered. It has never meant true." A fabricated poll stored on a perfectly secured server, she noted, passes every integrity check a security program runs.
From open-source champion to cautionary voice
O'Brien said she spent years as an advocate for bringing publicly available information into classified analysis. In Kabul in 2014, after an explosion near the base, she used a Twitter feed to locate the vehicle, the intersection and early casualty estimates while official reporting channels still had nothing.
That same openness, she said, has since been turned against the United States. During the recent escalation with Iran, she cited research from Cyabra, a firm she advises, identifying 37,000 pieces of AI-generated or manipulated content in the first week, drawing 145 million views. "It wasn't random," she said. The content pushed specific narratives aimed at Pentagon leaders, industry decision-makers and the public.
She described three intents behind campaigns like this:
- Gray-zone signaling. "We're not at peace, we're not at war." Adversaries demonstrate access to U.S. critical infrastructure, such as water systems, to send a message without striking.
- Division. Amplifying the extremes of domestic debates on politics, vaccines and the economy to distract and divide.
- Economic effect. Manufactured sentiment moves markets. In August, fabricated polls from an unknown firm were reposted by a mayoral campaign and moved prediction-market odds within minutes, before anyone checked whether the polls were real.
A blind spot for CISOs
"Organizations are being targeted in ways that fall outside of the traditional scope of what a security organization is responsible for," said Parham Eftekhari, EVP of Communities at CyberRisk Alliance (the parent company of SC Media) and Chairman of the Institute for Critical Infrastructure Technology, who joined O'Brien on stage. In most organizations, security teams that were already struggling with basic cyber hygiene, such as IAM, patching, security awareness training and data protection, are now consumed by reducing AI-related risk.
Eftekhari went on to say that the use of poisoned external data and misinformation to cause harm or fraud is something that rarely comes up in his work with CISOs. Behind every major brand decision, he said, "it's just people," and people can be influenced. AI, he added, is making that influence faster and more convincing.
The boardroom case study
To make the business risk concrete, O'Brien walked through the August 2025 backlash to Cracker Barrel's logo change, using data from Cyabra. Of roughly 3,200 profiles driving the conversation on X, 21% were fake. Those accounts produced more than 1,000 pieces of content that reached 4.45 million potential views.
The company's stock decline coincided with that amplification, she noted, without claiming it caused it. The board ultimately reversed the decision.
The case, she said, comes down to three questions: for every external source a consequential decision depends on who owns it, who verified it, and who in the room would know if it were wrong.
Eftekhari pressed the dilemma further: even with the data in hand, would leadership have acted differently? "Perception is reality," he said. That is where war-gaming earns its place, he argued: rehearse a brand decision under a manufactured backlash before it happens, not during.
The takeaway
O'Brien closed on the same three questions: for every external source a consequential decision depends on who owns it, who verified it, and who in the room would know if it were wrong. The warning sign, she said, is internal and external data that disagree. Most organizations, she added, have no formal process for the external data now flowing into decisions.
O'Brien contrasted the EU, where the AI Act has required labeling of AI-generated content and disclosure by chatbots since August, with the United States, which closed its State Department counter-disinformation office and sidelined CISA's foreign-influence staff in 2025.
