Estimated reading time: 6 minutes
Workplace AI adoption is moving faster than any technology before it. By early 2026, 43% of U.S. workers were using AI on the job. Most cyber programs were not built for that speed. The internal cybersecurity exposure this creates is called shadow AI. It is now the headline risk in Travelers’ Q1 2026 Cyber Threat Report.
The report carries a blunt subtitle. It names the moment “The Growing AI Governance Gap.” Lauren Winchester leads Cyber Risk Services at Travelers. She puts the danger in plain terms. The biggest near-term AI exposure is not the threat actor. It is the employee, “employee use of unsanctioned AI tools and apps – remains a primary threat”, said Winchester.
Shadow AI Is the Threat Already Inside the Building
Speculation about AI-powered attack chains keeps growing. The evidence points the other way. Travelers sees the documented risk inside the network, not outside it.
Shadow AI means employees’ use of unsanctioned tools and apps. The adoption numbers explain the urgency. AI reached 43% workplace use faster than personal computers or the internet did at the same stage.
A hands-off policy does not remove the exposure. It hides it. “Organizations with no official AI adoption policy don’t avoid the problem; they just lose visibility into it,” Winchester said. Staff export data to personal devices. They run unapproved tools off the network. They conceal the activity because no permitted options exist.
Real losses are starting to surface. These breaches are hard to detect. Winchester is candid about where this stands. The trend is not frequent or severe yet. “We know it is brewing,” she said.
For underwriters, this is a visibility problem. An insured that cannot see its own AI use cannot answer application questions accurately. That raises disclosure and controls questions at renewal. AI governance is starting to look like the next controls-underwriting frontier. MFA filled that role in 2021. Travelers folds this guidance into its Cyber Risk Services offering, which now ships with every cyber policy.
Ransomware Holds Near Record Levels
The threat backdrop stayed severe. Leak sites recorded 2,405 victims in Q1 2026. That figure sits just 2% below the all-time high set in Q4 2025. And it runs 7% above the same quarter last year.
Fragmentation defined the quarter. A record 84 ransomware groups were active. Nineteen of them appeared on leak sites for the first time. A new group called Gentlemen ranked second by volume. It posted 207 victims within months of emerging.
A crowded field is harder to police. Taking down one dominant group means less when dozens remain. The barrier to entry keeps falling. Ransomware-as-a-service supplies the tools.
When the Negotiator Is a Bot
One AI shift is already reaching the claim file. Travelers’ negotiation specialists have seen bots replace human attackers in ransom talks. The clue is speed. Data vanished from a leak site the instant it was prompted.
“That instantaneous response is the tell,” Winchester said. A bot behaves differently from a tired human juggling many cases. “You can’t tire out a bot. A bot instructed not to accept below a certain threshold will not waver,” she said. A bot told to hold holds.
This changes tactics more than strategy. A foe you cannot wear down removes one lever. Value then shifts toward the levers an insured controls. That means backup posture. It means recovery capability. And that means the option to avoid paying at all. Bots are not flawless. Far from it. Travelers is watching for new weaknesses.
The underwriting message is unambiguous. Pre-loss controls matter more as the negotiation lever weakens.
Social Engineering Fraud Goes Multi-Vector
The basic playbook has not changed much. Phishing, vishing, and business email compromise remain core tactics. The damage remains high. Social engineering fraud and BEC make up roughly 40% to 50% of Travelers’ cyber claims. Their combined severity is up more than 30% since 2023.
So is ransomware getting too much attention? Winchester says no. “Ransomware claims are typically much more severe than SEF and BEC,” she said. Encryption drives disruption and cost. Social engineering still deserves focus. It is often preventable with good hygiene. That makes it the insurer’s job to keep flagging it.
The tactics are now stitched together. The report details a “mail bomb plus ClickFix” pattern. Attackers flood an inbox to create panic. A fake IT contact then offers to help. The victim pastes a malicious command and hands over access. None of the steps looks like classic phishing. “The user feels like a participant in their own rescue,” Winchester said, adding that it “makes these attacks harder to train against.”
What It Means for Underwriters and Brokers
Two of the report’s data points sit in tension. Leak-site activity has tripled since 2022. Travelers’ own ransomware claims rose 80% over the same period. Those curves do not match.
The gap is easy to misread. Leak-site postings track the wider criminal ecosystem. Travelers’ claims track one insured book. That book has faced years of tightening controls. Slower claim growth than the open market may signal that controls underwriting is working. It is not proof of mispricing.
Winchester’s remit is risk mitigation, not pricing. Her team’s guidance points to actions an insured can take now. That follows the prevention-first message Travelers has carried in prior interviews with Cyber Insurance News. The Q1 report lists a clear governance starting kit. Name an accountable owner. Publish an acceptable use policy. Train every user. Keep a human in the loop for important decisions. Vet tools before deployment.
The through-line is simple. The 2026 risk picture rewards work done before a loss. Governance, training, and controls now carry the weight that perimeter defense once held alone. The Q1 report argues that the most useful response needs to start at home.
FAQ – Shadow AI And Cyber Insurance
Shadow AI is employee use of AI tools and apps that the organization has not sanctioned. It often happens off the network and without leadership visibility.
Because the evidence is documented today. AI-powered attack chains remain mostly a forecast. Confidential data is already turning up in unauthorized AI tools.
Not yet. They are emerging. The giveaway is instant response speed, such as data removed from a leak site the moment it is prompted.
Leak sites recorded 2,405 victims. That is near the all-time high. A record 84 groups were active, and a new group called Gentlemen ranked second by volume.
Name an accountable AI owner. Publish an acceptable use policy. Train users. Keep a human in the loop for high-stakes decisions. Vet tools before deployment.
Related Cyber Insurance Posts
- Most CISOs Are Fighting The Last War. AI-Powered Cyber Attacks Just Changed The Rules – The latest Cyber Insurance news PODCAST
- Small Business Cyber Insurance Adoption Up 50%, Insureon Data Shows
- The Call Is Coming from Inside the Office: Tackling Insider Cybersecurity Threats(Opens in a new browser tab)
- Cyber Risk Tops Business Concerns as Threats Surge, Travelers Report Finds(Opens in a new browser tab)
- Cybercrime Trends Report Shows Fraudsters Shift To Browsers, Bots, And Synthetic Identities(Opens in a new browser tab)