BLUF: On September 28, 2026, voice-security vendor Pindrop published its 2026 Deepfake Readiness Index. It reports that 74% of surveyed US security leaders encountered or suspected a deepfake attack in the past year, yet only 10% have tools built for the threat. The survey is vendor-run and small, so treat the figures as directional. The pattern it describes, concern without action, matches the incidents we see reported. The practical fix is mostly procedural and costs little.
What Is Reported
The account below follows Pindrop's release and trade coverage including Biometric Update, published September 28-30.
- Method. Wakefield Research surveyed 250 US security leaders at organizations with 1,000 or more employees, June 11-22, 2026. Pindrop states a margin of error of plus or minus 6.2 points at 95% confidence.
- Exposure. 74% said their organization had encountered or suspected a deepfake attack in the past year.
- Cost. Among organizations that experienced or suspected an attack, nearly half reported total costs of $500,000 or more, and one in four reported $1 million or more, counting direct losses, remediation and staff time. 49% also faced a follow-on cyberattack such as ransomware.
- Readiness. 93% said they are concerned their organization is not prepared. Only 10% reported purpose-built deepfake tools, and 37% rely on legacy defenses.
- Urgency. 75% said it will take a company leader being personally fooled or impersonated before deepfakes become a boardroom priority. 37% believe a single attack could put a company like theirs out of business.
What the Survey Can and Cannot Tell You
Pindrop sells deepfake detection. That does not make the numbers false, but it shapes which question gets asked and which conclusion is drawn. Three limits matter.
"Encountered or suspected" is a soft measure. The 74% combines confirmed attacks with suspected ones. It shows how common the worry is more than how common confirmed loss is.
The sample is narrow. 250 respondents, all US, all at large enterprises, all security leaders. It says little about smaller firms, family offices or individuals, who often have fewer controls.
Costs are self-reported. Staff time and remediation are hard to price consistently, and the survey does not say how respondents calculated them.
The finding with the most substance is the gap itself: near-universal concern alongside low adoption of dedicated tools. That is a behavioral observation, and it fits a pattern we have noted in reported incidents such as the Fideuram voice-cloning case, where the weakness was an approval path that trusted a familiar voice.
Why Concern Does Not Turn Into Action
The most revealing line in the data is the 75% who say a leader must be personally fooled first. It describes a waiting strategy: the organization will act once the cost is felt at the top.
There is also a quieter optimism. 74% of respondents believe defenses will improve faster than attack quality. That belief is unproven, and the incident record so far does not support leaning on it.
Both attitudes put the burden on a future tool or a future incident. Neither asks what the organization can change this quarter.
What to Change Without Buying Anything
Detection tools have a place. They are not the first step, and a tool without a procedure fails the same way a voice does: someone trusts the output and moves on. Start with how decisions get approved.
Separate the channel that asks from the channel that confirms. Any request for money, credentials or access that arrives by voice, video or message is confirmed through a different channel, using a contact detail you already hold. Not a number or link supplied in the request.
Set a callback rule with no seniority exemption. Impersonators choose the most senior name available. A rule that executives can skip is the rule attackers will use.
Agree a verification phrase for principals. A short, offline-agreed code between an executive and their assistant, driver or finance lead gives staff something a clone cannot guess. Rotate it and never send it digitally.
Reduce the raw material. Voice cloning needs a sample, often public. Review the interviews, conference recordings and social video that exist of your principals. Our note on executive digital footprint reduction covers the method.
Rehearse once. Run an internal drill in which a finance or operations lead receives a convincing request. Measure whether the callback rule held, not whether people spotted the fake. Our overview of deepfake detection services sets out where tooling fits after these steps.
Confidence and Caveats
High confidence that the survey was published and says what is described above, since we read the vendor release and independent trade coverage of it. Moderate confidence that the direction is right: other recent reporting on voice-cloning fraud points the same way. Low confidence in the precise percentages as a measure of the wider market, given a single vendor-commissioned survey of 250 respondents. We have not audited the underlying data.
Frequently Asked Questions
What did Pindrop's 2026 Deepfake Readiness Index find? That 74% of surveyed US enterprise security leaders encountered or suspected a deepfake attack in the past year, and 10% have purpose-built tools to address it.
Is the survey reliable? It is a credible directional signal from a named research firm, but it is vendor-commissioned, covers 250 US respondents, and uses a soft "encountered or suspected" measure. Read it as direction, not market size.
What is the cheapest control against a deepfake request? Confirming any request for money or access through a second, independently known channel, with no exemption for senior names.
Do organizations need detection software? Possibly, for high-volume channels such as call centers. For executive impersonation, a verification procedure is the first control and tooling supports it.
Kronus Intelligence Group builds and operates custom intelligence infrastructure for organizations that cannot afford to be wrong. If your environment requires it, we want to hear from you.
Start a Project →