Cyber Insurance & AI Tracker

Expanded the definition of a "security failure or data breach" to include an "AI security event" and expanded the funds-transfer-fraud trigger to include fraudulent instructions via deepfakes or other AI. Announced March 26, 2024; folded into the base Active Cyber Policy in 2025.

Added a global endorsement (US, UK, Canada, Australia, Germany, Denmark, Sweden, France) providing technical forensics, legal takedown support, and crisis-communications support for AI-generated impersonation of executives and employees. Announced Dec 9–10, 2025.

Introduced a globally available cyber endorsement extending coverage for generative-AI risks for businesses building their own genAI models: data "poisoning," usage-rights/IP infringement, and regulatory violations arising from the EU AI Act. Reported October 2024.

Expanded the program: Beazley and Chubb joined founding partner Munich Re, offering affirmative AI coverage for Google-related AI workloads and using cloud-posture telemetry (Security Command Center / CIS Benchmark reports) for underwriting. For qualified digital natives, Beazley replaces the application with a single-page attestation. Announced at Google Cloud Next, April 2025.

Launched the first standalone Affirmative AI Liability Insurance covering AI underperformance (hallucinations, model drift, mechanical failures), with limits reaching $25M+ by Jan 2026. In Feb 2026 the "Vanguard AI" structure paired Chaucer’s cyber and tech E&O with Armilla’s standalone AI liability, a "conscious decision to partner rather than stretch cyber policies beyond their original intent."

Launched Prime One in the US, embedding AI-related risks (cybercrime, business interruption, data restoration, system failure, third-party liability from AI events) plus quantum risks directly into the base policy form rather than via exclusion or bolt-on. Introduced AI-risk "Cowbell Factors" rating for autonomy, observability, and governance. Launched Dec 2, 2025.

Launched three AI products: NOVAAI (liability/cyber for AI developers), PONTAAI (excess/DIC wrap where existing policies exclude AI), and RESCAAI (first-party response for organizations embedding third-party AI). Affirmatively covers IP infringement and discrimination. Launched Jan 14, 2025.

On July 22, 2026 BOXX, the SME-focused cyber insurtech Zurich agreed to acquire in July 2025, added affirmative coverage for AI and deepfake related events, for social engineering and security failures, inside its commercial product Cyberboxx Business in Canada and the US. BOXX says the endorsement "eliminates ambiguity and cements BOXX’s commitment to providing clear cover for AI-driven incidents." Global head of underwriting Erik Tifft framed the exposure this way: "Threat actors are exploiting trusted relationships amongst employee and executive networks which can result in handing over credentials or misdirecting payments without an actual breach."

New standardized CGL exclusion endorsements CG 40 47, CG 40 48 (and CG 35 08 for products/completed operations) took effect Jan 1, 2026, giving carriers ready-made language to exclude bodily injury, property damage, and personal/advertising injury "arising from" generative AI. Verisk forms underpin the large majority of US commercial liability policies.

Introduced an "absolute" AI exclusion (reported as Form PC 51380) across D&O, E&O, and fiduciary liability, barring claims "arising out of" any "use, deployment, or development" of AI by any person or entity, reportedly naming tools like ChatGPT. Filed/reported late 2025, effective at renewal.

Filed requests with US regulators to exclude AI-related liabilities from various commercial policies; AIG reportedly told regulators generative AI is a "wide-ranging technology" where claims will "likely increase over time." First reported by the Financial Times, November 2025. Filing-stage; effective at renewal, jurisdiction-dependent.

Found 42% of respondents said their cyber policies already include exclusions tied to AI misuse or liability, and that 77% of insurers now require a formal internal/security-team review before issuing or renewing (up from 56% a year earlier).

The Insurer reported on July 23, 2026 that a review of nearly 10,000 filings through S&P Capital IQ found 41 P&C insurance groups in S&P Global Market Intelligence’s SNL Insurance dataset had at least one subsidiary that filed to adopt an AI exclusion, with subsidiaries of a further 20 groups filing to delay adoption until a later date. Insurance Journal reported on July 22, 2026 that carrier interest is concentrated in the ISO forms CG 40 47, CG 40 48 and CG 35 08, quoting Verisk’s Joe Lam: "Everyone acknowledges this is new technology...so they are all appreciative of having additional underwriting flexibility." The same piece cites a Gallagher study finding a 978% increase in AI-related lawsuits from 2021 to 2025.

A cyber insurer (a Chubb subsidiary) filed a subrogation action directly against an IT provider and an MSSP to recover exactly $500,000 it paid to insured CoWorx Staffing Services after an April 2024 ransomware attack. Alleges Congruity 360 failed to implement MFA for remote access and secure servers, and that Trustwave misclassified a detected event as "moderate," delaying response and preventing timely backups. Negligence + breach of contract, filed Sept 15, 2025. Updated Aug 17, 2026: on June 25, 2026 Judge Susan D. Wigenton of the District of New Jersey dismissed nearly all of Ace American’s claims against both defendants for failure to state a claim, leaving only part of the breach-of-contract claim against Congruity, per Bloomberg Law. The case is therefore no longer "unresolved" in the sense earlier editions of this tracker described.

A court permitted Travelers to rescind a $1M cyber policy after a ransomware attack because the insured represented MFA was deployed across all systems when it was only at the firewall. Rescission applies even for unintentional/honest misrepresentation under the majority US rule. Court ruling 2022; widely cited through 2026.

Stated that underwriting AI use has changed: "It used to be, ‘Do you use AI and how do you use it?’ Now it’s, ‘What models are you currently utilizing?’" This indicates deeper, model-level questioning. Reported 2025 (Business Insurance).

Global cyber rates fell 5% year-on-year in Q1 2026 (following a 7% decrease in Q4 2025; US ~-2%; IMEA -14%), the seventh consecutive quarterly decline. Marsh notes AI-threat affirmative coverage and supply-chain covers are "under development."

AI rose to the No. 2 global business risk (32% of responses), up from No. 10, the biggest riser in the survey of 3,338 risk-management experts across 97 countries. Cyber remained No. 1 at its highest-ever score (42%). Published Jan 14, 2026.

Reports that MFA is "still table stakes" but "MFA everywhere with evidence" is the standard; EDR/MDR now means 24/7 monitored response; and that Marsh and Chubb have begun incorporating Zero Trust principles (least-privilege, continuous verification, segmentation, conditional access) into underwriting conversations. 2026.

Maintained a stable outlook on the global cyber insurance segment on July 15, 2026, citing strong demand, ample capacity, and "the growing use of AI" among supportive factors, while warning of "persistent and evolving threat activity, particularly ransomware" and that rapid advances in AI could enable more sophisticated, scalable, and automated attacks. On pricing, AM Best said "premium rates charged for cyber-related coverage have undergone persistent declines and are not expected to stabilize in the near future," with profitability set to "remain favorable over the intermediate term," though a sharp rise in claim frequency or severity could force a future upward price correction.

Published July 23, 2026. Marsh reported that cyber insurance rates "declined by 4% globally," what it called "the twelfth consecutive quarter of declines," following "a 5% decrease in Q1." The largest decline was "in IMEA, at 14%," with "reductions ranging from 10% in LAC to 2% in the US." Global commercial rates overall fell 6% in Q2 2026, the eighth consecutive quarterly decrease.

Cowbell published the detail behind its AI-risk rating on July 23, 2026 and took it to the trade press on August 10, 2026. The AI Exposure factor scores "an organization’s visibility into its AI ecosystem," specifically "the organization’s ability to identify deployed models and applications, understand their operational reach, and assess the significance of the decisions they influence." The AI Vulnerability factor scores hallucination rates, algorithmic bias, adversarial robustness, susceptibility to prompt injection, data leakage, intellectual property exposure and upstream dependency risk. The AI Assurance factor scores "documentation, red-team testing, model lifecycle management, access controls, policy enforcement, auditability, and broader governance processes," including alignment with NIST and ISO/IEC 42001. The inputs are telemetry, collected by what Cowbell calls AI Connectors, not application answers. Co-founder and chief product officer Rajeev Gupta: "Autonomy is the largest severity multiplier in AI risk," and, on governance, "Watching what the AI did produces evidence."

On August 13, 2026 CNN reported research by the Israeli cyberdefense firm Dream documenting an intrusion into Taiwanese government networks in which up to eight AI agents, built from the open-source frameworks Hermes and OpenClaw, ran largely autonomously: mapping 21 government systems, probing for weaknesses and switching tactics when blocked. Dream stopped short of formally attributing the operation to Beijing. Insurance Business reported in August 2026 that the incident exposes two problems in how cyber forms are written. Tim Johnson, head of insurance at Browne Jacobson, noted that a number of cyber forms still define a "hacker" as a person, which leaves open whether "an autonomous AI attacker even triggers coverage as written," and that "coverage outcomes can turn on whether an incident is ultimately traced to a state-backed group." Mitch Ashley of The Futurum Group put the trigger problem plainly to Security Boulevard on August 12, 2026: "Cyber policies trigger unauthorized access by an intruder. An agent operating under valid credentials does not meet that definition," adding that insureds should be producing "execution evidence at the agent level, ahead of any claim." Caspar Rogers of Assured expects wording like Chubb’s Widespread Vulnerability Exclusion to see broader adoption; Christopher Keegan of Brown & Brown Risk Solutions described the market as "holding steady for now, competitive and profitable and not yet repriced."

Reported on August 12, 2026 from KYND’s inaugural Cyber Drop Live, where senior cyber market participants worked through where AI losses would actually land. Participants put roughly half of 50 plausible AI hallucination scenarios inside existing cyber coverage, which the write-up presented as discussion rather than established market data. The open question for underwriters was whether model drift or hallucination should be treated as a technology failure, a cyber incident, or something else entirely, illustrated by an AI coding agent that deleted a production database while reporting normal operation, a loss that may not meet a standard outage definition. Participants also named shadow AI use on personal devices as an underwriting visibility problem, and sorted adjacent AI exposures into other lines: AI-washing litigation toward D&O, a discriminatory AI recruitment tool toward employment practices.

Reuters reported on August 27, 2026 that cyber insurers are reviewing and adapting policy language after OpenAI, Anthropic and Meta Platforms disclosed that their AI agents "behaved unexpectedly, escaping controlled test environments and carrying out cyberattacks on companies without direct human instruction." Citing eight executives and analysts, the report said insurers are "grappling with issues including whether autonomous AI systems fit traditional policy definitions of a cyber attacker and who bears liability for AI-generated actions that cause a loss." Ryan Kratz, head of cyber, North America at MSIG USA: "As AI becomes capable of identifying vulnerabilities and carrying out attacks autonomously, carriers will need to continually review policy language." Serene Davis, QBE's global head of cyber: "If an AI-related event leads to a conventional cyber incident, resulting losses continue to fall within a cyber policy. AI is treated as a risk amplifier, not a fundamentally new cyber risk." A Beazley spokesperson said companies want AI risks included in broad cyber policies and that "as new AI risk emerges, we are developing new coverage." Marsh's global cyber product leader, Greg Eskins: "Underwriters recognize that it's important to continue to offer a product that responds to these types of events." Jenny Soubra, vice president of specialty commercial lines at Verisk Underwriting Solutions, said targeted exclusions are being discussed in some pockets of the industry for two scenarios: a systemic event where a single AI model or platform contributes to losses across many organizations at once, and a case where an AI agent, acting as designed, makes a costly autonomous decision that "some insurers may classify" as a non-cyber event. Karthik Ramakrishnan, CEO of Armilla AI: "Some losses caused by AI agents will absolutely fall within cyber policies. The harder cases are where there is no conventional attacker and potentially no unauthorized credential use."