[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-emerging-enterprise-and-insurance-risks-from-rapid-ai-adoption-en":3,"ArticleBody_tyGI5hx56eH4zSpYavHDgl70s86f2NaQJfoluT0NPDs":219},{"article":4,"relatedArticles":190,"locale":66},{"id":5,"title":6,"slug":7,"content":8,"htmlContent":9,"excerpt":10,"category":11,"tags":12,"metaDescription":10,"wordCount":13,"readingTime":14,"publishedAt":15,"sources":16,"sourceCoverage":58,"transparency":60,"seo":63,"language":66,"featuredImage":67,"featuredImageCredit":68,"isFreeGeneration":72,"trendSlug":7,"trendSnapshot":73,"niche":80,"geoTakeaways":84,"geoFaq":93,"entities":103},"6a5768d9b14fe5915b3ecf3e","Emerging Enterprise and Insurance Risks from Rapid AI Adoption","emerging-enterprise-and-insurance-risks-from-rapid-ai-adoption","## 1. How Rapid AI Adoption Is Rewiring Enterprise Risk\n\nEnterprise AI has moved into production: operational use jumped from 45% to 63% in a year, yet fewer than half of organizations have a formal AI risk framework.[5] About 85% consume AI services (including LLMs and generative tools), while 25% do not know which systems are in use.[5][6] Adoption without visibility is where risk compounds.\n\nAgentic AI widens this gap. Task‑specific agents are now embedded into enterprise apps and can execute multi‑step workflows—procurement, deployment, customer outreach—at machine speed with limited oversight.[1] [Gartner](\u002Fentities\u002F693feb44312dc892c4c19011-gartner) expects around 40% of enterprise apps to include such agents by 2026, up from less than 5% in 2025.[1]\n\n- **Autonomy multiplies blast radius:** When agents orchestrate across [CI\u002FCD](\u002Fentities\u002F6984cfcbe28785d1e150d753-cicd), CRM, cloud, and finance tools, a single misconfiguration or compromise can trigger system‑wide events, not isolated incidents.[1]  \n- **Expanded privileges:** To be useful, agents get broad permissions—API keys, [SSO](\u002Fentities\u002F69886e7b033ff25c8c61b12a-sso), and write rights across SaaS and cloud. Compromise can enable privilege escalation, outages, or mass data exfiltration in one automated run.[1]\n\n[Shadow AI](\u002Fentities\u002F6984ceb6e28785d1e150d5f3-shadow-ai) has become a persistent data‑leak vector. Employees paste contracts, strategy decks, and production code into public chatbots to “speed up” work, bypassing [DLP](\u002Fentities\u002F6984ceb6e28785d1e150d5f2-dlp) and logging.[7] Once data enters third‑party LLMs, enterprises lose control over storage, training use, and onward sharing.[7][10]\n\nA security lead at a 30‑person SaaS firm found a junior developer had pasted live logs—including access tokens—into a public chatbot for months, none of it visible in SIEM or [CASB](\u002Fentities\u002F6a0a83e41f0b27c1f426ad6c-casb) tools yet materially expanding breach surface.[10]\n\n📊 **Data:** [ChatGPT](\u002Fentities\u002F6939891c312dc892c4c183ff-chatgpt) alone triggered over 410 million DLP policy violations in 2025, each a sensitive‑data egress attempt via AI tools.[10] Shadow AI incidents are growing while 97% of companies lack robust AI access controls and 63% have no AI policies, turning routine productivity habits into systemic data, privacy, and regulatory threats.[8]\n\n## 2. Specific Risk Categories Driving Insurance Demand\n\nAI risk needs a taxonomy that goes beyond generic “cyber.” Key categories include:\n\n- Data leakage and privacy violations  \n- IP and trade‑secret exposure  \n- [Model and content liability](https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FModel_release) (hallucinations, defamation, copyright)  \n- Operational disruption from autonomous agents  \n- Strategic risk from over‑reliance on immature models and vendors[2]\n\nFinancial and regulatory impacts are already material:\n\n- AI without proper governance raises average breach costs by ~\\$670,000 versus governed environments.[6]  \n- Global data‑ and AI‑related fines topped \\$10 billion in 2023, and roughly half of governments are expected to enforce AI‑specific obligations (e.g., EU AI Act, GDPR‑aligned rules) by mid‑2026.[6]\n\n💡 **Key takeaway:** For boards, the dominant AI cost is often regulatory drag, remediation, and re‑engineering after unmanaged failures, not the tools themselves.[6]\n\nShadow AI is now a distinct insurable exposure:\n\n- Core risk: regulated or high‑value data sent to external generative AI endpoints the enterprise cannot fully contract, audit, or control.[7][8][10]  \n- Open questions on prompt logging, retention, and training use directly affect privacy liability and long‑tail IP risk.\n\nThe Geneva Association notes that generative AI is being embedded into products, underwriting, and customer channels, creating new liability paths:\n\n- Model error and AI‑driven discrimination  \n- Content‑related claims (e.g., harmful or infringing outputs)[2]\n\nTraditional cyber, PI, and D&O wordings rarely address these explicitly.[2]\n\nInsurers face a dual balance‑sheet exposure:\n\n- They are now close to tech and media firms in predictive and generative AI adoption and outpace most other sectors.[3]  \n- Generative AI investments are expected to grow by >300% from 2023 to 2025 as carriers move from pilots to scale.[4]\n\nAI failures can hit:\n\n- **Operational resilience:** claims, pricing, customer service  \n- **Underwriting portfolios:** systemic AI incidents across insureds[2][3][4]\n\n💼 **Implication:** Insurers are both heavy AI users and risk aggregators; their own AI governance is a solvency issue, not just an efficiency project.[3][4]\n\n## 3. Governance, Controls, and Insurance Readiness\n\nThe evidence is consistent: the real cost is the absence of governance. Ad hoc AI use correlates with higher breach losses and expensive retrofits.[6] Organizations with comprehensive AI governance achieve roughly 30% better ROI on AI portfolios, turning compliance into an accelerator.[6][9]\n\nEffective AI governance frameworks typically include:\n\n- Clear accountability and decision rights for AI  \n- Central inventory of models, agents, and embedded AI features  \n- Use‑case approvals and risk assessments by data type and context  \n- Continuous monitoring of performance, drift, and fairness  \n- Transparent reporting on AI health, incidents, and value[8][9]\n\n⚡ **Control theme:** Governance must be adaptive and near real‑time to match shifting model behavior and autonomous agent actions.[9]\n\nFor underwriters, governance translates into concrete controls, such as:\n\n- Shadow AI inventories and classification (sanctioned vs. unsanctioned)  \n- DLP tuned to AI endpoints with prompt\u002Ffile inspection  \n- Zero‑trust and least‑privilege access for agents and service accounts  \n- Prompt‑injection defenses plus human‑in‑the‑loop for high‑impact workflows  \n- Data‑centric protections blocking sensitive content from public tools[1][7][10]\n\nInsurer responses can include:\n\n- AI‑specific endorsements and exclusions within cyber and PI  \n- Dedicated covers for model\u002Falgorithm failure, including bias or hallucination‑driven loss  \n- Explicit wording on training‑data use, retention, and shared models  \n- Pre‑bind risk surveys benchmarking AI maturity against peers and best practice[2][5][9]\n\n⚠️ **Key point:** Pricing AI risk without insight into governance maturity is akin to pricing cyber without knowing if basic patching exists.\n\n## Conclusion: From Emerging Threat to Shared Playbook\n\nAgentic systems, Shadow AI, and governance gaps are reshaping enterprise risk faster than traditional cyber frameworks and policy wordings can adapt.[1][2][6] For insurers, this is both acute exposure and a strategic opening to lead on responsible AI design, deployment, governance, and disclosure.[2][3]\n\nRisk leaders and insurance executives should now jointly:\n\n- Inventory current AI, agents, and Shadow AI usage  \n- Assess governance maturity against frameworks such as the NIST AI RMF and ISO 42001  \n- Engage brokers and underwriters with an evidence‑based view of AI risk posture","\u003Ch2>1. How Rapid AI Adoption Is Rewiring Enterprise Risk\u003C\u002Fh2>\n\u003Cp>Enterprise AI has moved into production: operational use jumped from 45% to 63% in a year, yet fewer than half of organizations have a formal AI risk framework.\u003Ca href=\"#source-5\" class=\"citation-link\" title=\"View source [5]\">[5]\u003C\u002Fa> About 85% consume AI services (including LLMs and generative tools), while 25% do not know which systems are in use.\u003Ca href=\"#source-5\" class=\"citation-link\" title=\"View source [5]\">[5]\u003C\u002Fa>\u003Ca href=\"#source-6\" class=\"citation-link\" title=\"View source [6]\">[6]\u003C\u002Fa> Adoption without visibility is where risk compounds.\u003C\u002Fp>\n\u003Cp>Agentic AI widens this gap. Task‑specific agents are now embedded into enterprise apps and can execute multi‑step workflows—procurement, deployment, customer outreach—at machine speed with limited oversight.\u003Ca href=\"#source-1\" class=\"citation-link\" title=\"View source [1]\">[1]\u003C\u002Fa> \u003Ca href=\"\u002Fentities\u002F693feb44312dc892c4c19011-gartner\">Gartner\u003C\u002Fa> expects around 40% of enterprise apps to include such agents by 2026, up from less than 5% in 2025.\u003Ca href=\"#source-1\" class=\"citation-link\" title=\"View source [1]\">[1]\u003C\u002Fa>\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>Autonomy multiplies blast radius:\u003C\u002Fstrong> When agents orchestrate across \u003Ca href=\"\u002Fentities\u002F6984cfcbe28785d1e150d753-cicd\">CI\u002FCD\u003C\u002Fa>, CRM, cloud, and finance tools, a single misconfiguration or compromise can trigger system‑wide events, not isolated incidents.\u003Ca href=\"#source-1\" class=\"citation-link\" title=\"View source [1]\">[1]\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Expanded privileges:\u003C\u002Fstrong> To be useful, agents get broad permissions—API keys, \u003Ca href=\"\u002Fentities\u002F69886e7b033ff25c8c61b12a-sso\">SSO\u003C\u002Fa>, and write rights across SaaS and cloud. Compromise can enable privilege escalation, outages, or mass data exfiltration in one automated run.\u003Ca href=\"#source-1\" class=\"citation-link\" title=\"View source [1]\">[1]\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Ca href=\"\u002Fentities\u002F6984ceb6e28785d1e150d5f3-shadow-ai\">Shadow AI\u003C\u002Fa> has become a persistent data‑leak vector. Employees paste contracts, strategy decks, and production code into public chatbots to “speed up” work, bypassing \u003Ca href=\"\u002Fentities\u002F6984ceb6e28785d1e150d5f2-dlp\">DLP\u003C\u002Fa> and logging.\u003Ca href=\"#source-7\" class=\"citation-link\" title=\"View source [7]\">[7]\u003C\u002Fa> Once data enters third‑party LLMs, enterprises lose control over storage, training use, and onward sharing.\u003Ca href=\"#source-7\" class=\"citation-link\" title=\"View source [7]\">[7]\u003C\u002Fa>\u003Ca href=\"#source-10\" class=\"citation-link\" title=\"View source [10]\">[10]\u003C\u002Fa>\u003C\u002Fp>\n\u003Cp>A security lead at a 30‑person SaaS firm found a junior developer had pasted live logs—including access tokens—into a public chatbot for months, none of it visible in SIEM or \u003Ca href=\"\u002Fentities\u002F6a0a83e41f0b27c1f426ad6c-casb\">CASB\u003C\u002Fa> tools yet materially expanding breach surface.\u003Ca href=\"#source-10\" class=\"citation-link\" title=\"View source [10]\">[10]\u003C\u002Fa>\u003C\u002Fp>\n\u003Cp>📊 \u003Cstrong>Data:\u003C\u002Fstrong> \u003Ca href=\"\u002Fentities\u002F6939891c312dc892c4c183ff-chatgpt\">ChatGPT\u003C\u002Fa> alone triggered over 410 million DLP policy violations in 2025, each a sensitive‑data egress attempt via AI tools.\u003Ca href=\"#source-10\" class=\"citation-link\" title=\"View source [10]\">[10]\u003C\u002Fa> Shadow AI incidents are growing while 97% of companies lack robust AI access controls and 63% have no AI policies, turning routine productivity habits into systemic data, privacy, and regulatory threats.\u003Ca href=\"#source-8\" class=\"citation-link\" title=\"View source [8]\">[8]\u003C\u002Fa>\u003C\u002Fp>\n\u003Ch2>2. Specific Risk Categories Driving Insurance Demand\u003C\u002Fh2>\n\u003Cp>AI risk needs a taxonomy that goes beyond generic “cyber.” Key categories include:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>Data leakage and privacy violations\u003C\u002Fli>\n\u003Cli>IP and trade‑secret exposure\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FModel_release\" class=\"wiki-link\" target=\"_blank\" rel=\"noopener\">Model and content liability\u003C\u002Fa> (hallucinations, defamation, copyright)\u003C\u002Fli>\n\u003Cli>Operational disruption from autonomous agents\u003C\u002Fli>\n\u003Cli>Strategic risk from over‑reliance on immature models and vendors\u003Ca href=\"#source-2\" class=\"citation-link\" title=\"View source [2]\">[2]\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>Financial and regulatory impacts are already material:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>AI without proper governance raises average breach costs by ~$670,000 versus governed environments.\u003Ca href=\"#source-6\" class=\"citation-link\" title=\"View source [6]\">[6]\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>Global data‑ and AI‑related fines topped $10 billion in 2023, and roughly half of governments are expected to enforce AI‑specific obligations (e.g., EU AI Act, GDPR‑aligned rules) by mid‑2026.\u003Ca href=\"#source-6\" class=\"citation-link\" title=\"View source [6]\">[6]\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>💡 \u003Cstrong>Key takeaway:\u003C\u002Fstrong> For boards, the dominant AI cost is often regulatory drag, remediation, and re‑engineering after unmanaged failures, not the tools themselves.\u003Ca href=\"#source-6\" class=\"citation-link\" title=\"View source [6]\">[6]\u003C\u002Fa>\u003C\u002Fp>\n\u003Cp>Shadow AI is now a distinct insurable exposure:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>Core risk: regulated or high‑value data sent to external generative AI endpoints the enterprise cannot fully contract, audit, or control.\u003Ca href=\"#source-7\" class=\"citation-link\" title=\"View source [7]\">[7]\u003C\u002Fa>\u003Ca href=\"#source-8\" class=\"citation-link\" title=\"View source [8]\">[8]\u003C\u002Fa>\u003Ca href=\"#source-10\" class=\"citation-link\" title=\"View source [10]\">[10]\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>Open questions on prompt logging, retention, and training use directly affect privacy liability and long‑tail IP risk.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>The Geneva Association notes that generative AI is being embedded into products, underwriting, and customer channels, creating new liability paths:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>Model error and AI‑driven discrimination\u003C\u002Fli>\n\u003Cli>Content‑related claims (e.g., harmful or infringing outputs)\u003Ca href=\"#source-2\" class=\"citation-link\" title=\"View source [2]\">[2]\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>Traditional cyber, PI, and D&O wordings rarely address these explicitly.\u003Ca href=\"#source-2\" class=\"citation-link\" title=\"View source [2]\">[2]\u003C\u002Fa>\u003C\u002Fp>\n\u003Cp>Insurers face a dual balance‑sheet exposure:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>They are now close to tech and media firms in predictive and generative AI adoption and outpace most other sectors.\u003Ca href=\"#source-3\" class=\"citation-link\" title=\"View source [3]\">[3]\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>Generative AI investments are expected to grow by >300% from 2023 to 2025 as carriers move from pilots to scale.\u003Ca href=\"#source-4\" class=\"citation-link\" title=\"View source [4]\">[4]\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>AI failures can hit:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>Operational resilience:\u003C\u002Fstrong> claims, pricing, customer service\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Underwriting portfolios:\u003C\u002Fstrong> systemic AI incidents across insureds\u003Ca href=\"#source-2\" class=\"citation-link\" title=\"View source [2]\">[2]\u003C\u002Fa>\u003Ca href=\"#source-3\" class=\"citation-link\" title=\"View source [3]\">[3]\u003C\u002Fa>\u003Ca href=\"#source-4\" class=\"citation-link\" title=\"View source [4]\">[4]\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>💼 \u003Cstrong>Implication:\u003C\u002Fstrong> Insurers are both heavy AI users and risk aggregators; their own AI governance is a solvency issue, not just an efficiency project.\u003Ca href=\"#source-3\" class=\"citation-link\" title=\"View source [3]\">[3]\u003C\u002Fa>\u003Ca href=\"#source-4\" class=\"citation-link\" title=\"View source [4]\">[4]\u003C\u002Fa>\u003C\u002Fp>\n\u003Ch2>3. Governance, Controls, and Insurance Readiness\u003C\u002Fh2>\n\u003Cp>The evidence is consistent: the real cost is the absence of governance. Ad hoc AI use correlates with higher breach losses and expensive retrofits.\u003Ca href=\"#source-6\" class=\"citation-link\" title=\"View source [6]\">[6]\u003C\u002Fa> Organizations with comprehensive AI governance achieve roughly 30% better ROI on AI portfolios, turning compliance into an accelerator.\u003Ca href=\"#source-6\" class=\"citation-link\" title=\"View source [6]\">[6]\u003C\u002Fa>\u003Ca href=\"#source-9\" class=\"citation-link\" title=\"View source [9]\">[9]\u003C\u002Fa>\u003C\u002Fp>\n\u003Cp>Effective AI governance frameworks typically include:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>Clear accountability and decision rights for AI\u003C\u002Fli>\n\u003Cli>Central inventory of models, agents, and embedded AI features\u003C\u002Fli>\n\u003Cli>Use‑case approvals and risk assessments by data type and context\u003C\u002Fli>\n\u003Cli>Continuous monitoring of performance, drift, and fairness\u003C\u002Fli>\n\u003Cli>Transparent reporting on AI health, incidents, and value\u003Ca href=\"#source-8\" class=\"citation-link\" title=\"View source [8]\">[8]\u003C\u002Fa>\u003Ca href=\"#source-9\" class=\"citation-link\" title=\"View source [9]\">[9]\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>⚡ \u003Cstrong>Control theme:\u003C\u002Fstrong> Governance must be adaptive and near real‑time to match shifting model behavior and autonomous agent actions.\u003Ca href=\"#source-9\" class=\"citation-link\" title=\"View source [9]\">[9]\u003C\u002Fa>\u003C\u002Fp>\n\u003Cp>For underwriters, governance translates into concrete controls, such as:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>Shadow AI inventories and classification (sanctioned vs. unsanctioned)\u003C\u002Fli>\n\u003Cli>DLP tuned to AI endpoints with prompt\u002Ffile inspection\u003C\u002Fli>\n\u003Cli>Zero‑trust and least‑privilege access for agents and service accounts\u003C\u002Fli>\n\u003Cli>Prompt‑injection defenses plus human‑in‑the‑loop for high‑impact workflows\u003C\u002Fli>\n\u003Cli>Data‑centric protections blocking sensitive content from public tools\u003Ca href=\"#source-1\" class=\"citation-link\" title=\"View source [1]\">[1]\u003C\u002Fa>\u003Ca href=\"#source-7\" class=\"citation-link\" title=\"View source [7]\">[7]\u003C\u002Fa>\u003Ca href=\"#source-10\" class=\"citation-link\" title=\"View source [10]\">[10]\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>Insurer responses can include:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>AI‑specific endorsements and exclusions within cyber and PI\u003C\u002Fli>\n\u003Cli>Dedicated covers for model\u002Falgorithm failure, including bias or hallucination‑driven loss\u003C\u002Fli>\n\u003Cli>Explicit wording on training‑data use, retention, and shared models\u003C\u002Fli>\n\u003Cli>Pre‑bind risk surveys benchmarking AI maturity against peers and best practice\u003Ca href=\"#source-2\" class=\"citation-link\" title=\"View source [2]\">[2]\u003C\u002Fa>\u003Ca href=\"#source-5\" class=\"citation-link\" title=\"View source [5]\">[5]\u003C\u002Fa>\u003Ca href=\"#source-9\" class=\"citation-link\" title=\"View source [9]\">[9]\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>⚠️ \u003Cstrong>Key point:\u003C\u002Fstrong> Pricing AI risk without insight into governance maturity is akin to pricing cyber without knowing if basic patching exists.\u003C\u002Fp>\n\u003Ch2>Conclusion: From Emerging Threat to Shared Playbook\u003C\u002Fh2>\n\u003Cp>Agentic systems, Shadow AI, and governance gaps are reshaping enterprise risk faster than traditional cyber frameworks and policy wordings can adapt.\u003Ca href=\"#source-1\" class=\"citation-link\" title=\"View source [1]\">[1]\u003C\u002Fa>\u003Ca href=\"#source-2\" class=\"citation-link\" title=\"View source [2]\">[2]\u003C\u002Fa>\u003Ca href=\"#source-6\" class=\"citation-link\" title=\"View source [6]\">[6]\u003C\u002Fa> For insurers, this is both acute exposure and a strategic opening to lead on responsible AI design, deployment, governance, and disclosure.\u003Ca href=\"#source-2\" class=\"citation-link\" title=\"View source [2]\">[2]\u003C\u002Fa>\u003Ca href=\"#source-3\" class=\"citation-link\" title=\"View source [3]\">[3]\u003C\u002Fa>\u003C\u002Fp>\n\u003Cp>Risk leaders and insurance executives should now jointly:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>Inventory current AI, agents, and Shadow AI usage\u003C\u002Fli>\n\u003Cli>Assess governance maturity against frameworks such as the NIST AI RMF and ISO 42001\u003C\u002Fli>\n\u003Cli>Engage brokers and underwriters with an evidence‑based view of AI risk posture\u003C\u002Fli>\n\u003C\u002Ful>\n","1. How Rapid AI Adoption Is Rewiring Enterprise Risk\n\nEnterprise AI has moved into production: operational use jumped from 45% to 63% in a year, yet fewer than half of organizations have a formal AI r...","trend-radar",[],951,5,"2026-07-15T11:11:10.916Z",[17,22,26,30,34,38,42,46,50,54],{"title":18,"url":19,"summary":20,"type":21},"Emerging Enterprise Security Risks of AI","https:\u002F\u002Fwww.recordedfuture.com\u002Fresearch\u002Femerging-enterprise-security-risks-of-ai","Emerging Enterprise Security Risks of AI\n\nPUBLISHED ON 21 APR 2026\n\nInsikt Group®\n\n[Download report](https:\u002F\u002Fassets.recordedfuture.com\u002FExecutive-Insights\u002Feir-2026-0319.pdf \"Download report\")\n\n## Summa...","kb",{"title":23,"url":24,"summary":25,"type":21},"Gen AI Risks for Businesses: Exploring the role for insurance","https:\u002F\u002Fwww.genevaassociation.org\u002Fpublication\u002Fdigital-ai-transformation\u002Fgen-ai-risks-businesses-exploring-role-insurance-0","Gen AI Risks for Businesses: Exploring the role for insurance\n\n2 October 2025\n\nSUMMARY PDF - EN\nSubscribe to our Digital & AI Transformation communications\nSubscribe & Download\nDownload report\n\nSUMMAR...",{"title":27,"url":28,"summary":29,"type":21},"Insurance Leads in AI Adoption. Now It’s Time to Scale.","https:\u002F\u002Fwww.bcg.com\u002Fpublications\u002F2025\u002Finsurance-leads-ai-adoption-now-time-to-scale","Insurance Leads in AI Adoption. Now It’s Time to Scale.\n\nBy Joe Khoury, Daniel Martines, Christopher Freese, Jürgen Eckel, Malli Gupta, and Lara Najjar\n\nArticle September 04, 2025\n\nKey Takeaways\n\n- De...",{"title":31,"url":32,"summary":33,"type":21},"Generative AI in the insurance industry","https:\u002F\u002Fwww.ibm.com\u002Fthought-leadership\u002Finstitute-business-value\u002Fen-us\u002Freport\u002Finsurance-generative-ai","Savvy insurers will seize the gen AI opportunities that intrigue their customers\n\nFor the insurance industry, is generative AI more of a risk or an opportunity? Insurance CEOs are equally divided: 49%...",{"title":35,"url":36,"summary":37,"type":21},"How Leaders Navigate AI: 2026 AI Adoption and Risk Survey","https:\u002F\u002Fwww.linkedin.com\u002Fpulse\u002Fhow-leaders-navigate-ai-2026-adoption-risk-survey-gallagher-global-umqtc","How Leaders Navigate AI: 2026 AI Adoption and Risk Survey\n\nIt’s no surprise that this issue of Confidence Matters focuses on the transformative nature of AI in business.\n\nA clear shift is taking place...",{"title":39,"url":40,"summary":41,"type":21},"The Hidden Cost of Moving Fast Without Guardrails","https:\u002F\u002Fwww.mindstudio.ai\u002Fblog\u002Fcompliance-first-ai-enterprise-deployments","The Hidden Cost of Moving Fast Without Guardrails\n\nEnterprise AI adoption hit a critical inflection point in 2026. While 85% of organizations now use AI services, a staggering 25% don’t know what AI s...",{"title":43,"url":44,"summary":45,"type":21},"The Invisible AI Risk: When Employees Share Sensitive Information with AI Tools","https:\u002F\u002Fwww.sealpath.com\u002Fblog\u002Fsensitive-information-ai-risks\u002F","The Invisible AI Risk: When Employees Share Sensitive Information with AI Tools\n\nExecutive Summary:\n\n- The Unseen Risks of Generative AI at Work: While tools like ChatGPT, Claude, or Copilot have beco...",{"title":47,"url":48,"summary":49,"type":21},"How to Control Shadow AI Risks Before It Turns Into a Data Breach","https:\u002F\u002Fpenncomp.com\u002Fhow-to-control-shadow-ai-risks\u002F","How many employees in your organization are using AI tools without approval? What if someone pasted a confidential report into an AI chatbot to summarize it? Or did a developer share internal code wit...",{"title":51,"url":52,"summary":53,"type":21},"The Dual Role of AI Governance in Growth and Risk Mitigation","https:\u002F\u002Fwww.modelop.com\u002Fai-governance","The goal of effective AI Governance is to accelerate innovation and business growth by 1) increasing the efficiency and efficacy of an organization's AI initiatives and 2) mitigating potential risks w...",{"title":55,"url":56,"summary":57,"type":21},"Shadow AI Data Risk: Your 30-Day Containment Strategy","https:\u002F\u002Fwww.zscaler.com\u002Fblogs\u002Fproduct-insights\u002Fshadow-ai-data-risk-30-day-containment-strategy","Shadow AI Data Risk: Your 30-Day Containment Strategy\n\nOverview\n\nYour employees shared sensitive data with artificial intelligence (AI) tools today. They did it to work faster, solve problems, and mee...",{"totalSources":59},10,{"generationDuration":61,"kbQueriesCount":59,"confidenceScore":62,"sourcesCount":59},254292,100,{"metaTitle":64,"metaDescription":65},"Enterprise AI risks: managing emergent & insurance gaps","Skyrocketing AI use creates shadow AI, agentic threats, and insurance gaps. Get concise risk insights plus 3 mitigation steps to lower loss. Find out how.","en","https:\u002F\u002Fimages.unsplash.com\u002Fphoto-1637763723578-79a4ca9225f7?ixid=M3w4OTczNDl8MHwxfHNlYXJjaHwxfHxlbWVyZ2luZyUyMGVudGVycHJpc2UlMjBpbnN1cmFuY2UlMjByaXNrc3xlbnwxfDB8fHwxNzg0MTEzMzY5fDA&ixlib=rb-4.1.0&w=1200&h=630&fit=crop&crop=entropy&auto=format,compress&q=60",{"photographerName":69,"photographerUrl":70,"unsplashUrl":71},"Vlad Deep","https:\u002F\u002Funsplash.com\u002F@vladdeep?utm_source=coreprose&utm_medium=referral","https:\u002F\u002Funsplash.com\u002Fphotos\u002Fa-magnifying-glass-sitting-on-top-of-a-piece-of-paper-mCqi3MljC4E?utm_source=coreprose&utm_medium=referral",true,{"score":62,"type":74,"sourceCount":59,"topSourceDomains":75,"detectedAt":79,"mentionsLast7Days":59},"spiking",[76,77,78],"allianz.com","ey.com","newsroom.ibm.com","2026-07-14T17:01:56.618Z",{"key":81,"name":82,"nameEn":83},"ia","Intelligence Artificielle","Artificial Intelligence",[85,87,89,91],{"text":86},"Enterprise AI production use rose from 45% to 63% in one year, while fewer than half of organizations have a formal AI risk framework.",{"text":88},"About 85% of organizations consume external AI services and 25% do not know which AI systems are in use, creating pervasive visibility gaps; ChatGPT alone triggered over 410 million DLP policy violations in 2025.",{"text":90},"Gartner forecasts ~40% of enterprise applications will include task‑specific agents by 2026 (up from \u003C5% in 2025), multiplying blast radius when agents have broad API, SSO, and write privileges.",{"text":92},"97% of companies lack robust AI access controls and 63% have no AI policies; insurers are increasing generative AI investments by >300% (2023–2025) while facing aggregated underwriting exposures.",[94,97,100],{"question":95,"answer":96},"How does agentic AI increase enterprise risk?","Agentic AI increases enterprise risk by expanding the attack surface and automating cross‑system actions at machine speed, so a single compromise or misconfiguration can trigger multi‑system outages or mass data exfiltration. Agents typically require broad privileges—API keys, SSO, and write access across CI\u002FCD, CRM, cloud, and finance tools—enabling privilege escalation and automated lateral movement. Rapidly embedded agents in apps remove human gating for many workflows, and Gartner’s projection that ~40% of enterprise apps will include agents by 2026 means the frequency and scale of these automated failure modes will grow materially unless least‑privilege, prompt‑safeguards, and real‑time monitoring are enforced.",{"question":98,"answer":99},"What controls most effectively reduce Shadow AI exposure?","Effective controls combine visibility, prevention, and behavior change: maintain a central inventory of sanctioned AI endpoints, deploy DLP tuned for AI inputs (prompt\u002Ffile inspection and blocking of regulated data), and enforce single‑pane logging for prompts and responses. Complement technical controls with policy—clear acceptable use rules, mandatory approved tools, and user training—plus runtime protections like prompt‑injection defenses and human‑in‑the‑loop for high‑risk operations. These measures together prevent sensitive data from being sent to uncontrolled third‑party LLMs and reduce long‑tail privacy, IP, and regulatory liabilities.",{"question":101,"answer":102},"How should insurers underwrite and price AI risk today?","Insurers must underwrite AI risk based on governance maturity and specific controls rather than generic cyber labels: require pre‑bind surveys on model inventories, agent usage, DLP for AI endpoints, access‑control posture, and incident logging for prompts and training data. Pricing should reflect quantifiable gaps—organisations without AI policies or DLP tuned to AI endpoints exhibit materially higher breach costs (average incremental costs reported ~\\$670,000); insurers should use control‑based scoring, stress‑test for systemic agent failures, and offer AI‑specific endorsements or exclusions to align coverage with residual exposures. Continuous monitoring and periodic reassessments are essential as agents and models evolve.",[104,112,118,124,130,136,142,148,155,161,167,173,180,185],{"id":105,"name":106,"type":107,"confidence":108,"wikipediaUrl":109,"slug":110,"mentionCount":111},"693985b8312dc892c4c18386","agentic AI","concept",0.99,null,"693985b8312dc892c4c18386-agentic-ai",278,{"id":113,"name":114,"type":107,"confidence":108,"wikipediaUrl":115,"slug":116,"mentionCount":117},"69822206e28785d1e150b8aa","prompt injection","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FPrompt_injection","69822206e28785d1e150b8aa-prompt-injection",145,{"id":119,"name":120,"type":107,"confidence":108,"wikipediaUrl":121,"slug":122,"mentionCount":123},"6984ceb6e28785d1e150d5f3","Shadow AI","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FShadow_IT","6984ceb6e28785d1e150d5f3-shadow-ai",92,{"id":125,"name":126,"type":107,"confidence":127,"wikipediaUrl":109,"slug":128,"mentionCount":129},"696287f719d266277e150ef0","Enterprise AI",0.98,"696287f719d266277e150ef0-enterprise-ai",40,{"id":131,"name":132,"type":107,"confidence":127,"wikipediaUrl":133,"slug":134,"mentionCount":135},"6984cfcbe28785d1e150d753","CI\u002FCD","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FCI%2FCD","6984cfcbe28785d1e150d753-cicd",33,{"id":137,"name":138,"type":107,"confidence":139,"wikipediaUrl":109,"slug":140,"mentionCount":141},"698772c9033ff25c8c61a3bd","SIEM",0.97,"698772c9033ff25c8c61a3bd-siem",29,{"id":143,"name":144,"type":107,"confidence":127,"wikipediaUrl":145,"slug":146,"mentionCount":147},"6984ceb6e28785d1e150d5f2","DLP","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FDLP","6984ceb6e28785d1e150d5f2-dlp",15,{"id":149,"name":150,"type":107,"confidence":151,"wikipediaUrl":152,"slug":153,"mentionCount":154},"6a0a83e41f0b27c1f426ad6c","CASB",0.95,"https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FCASB","6a0a83e41f0b27c1f426ad6c-casb",8,{"id":156,"name":157,"type":107,"confidence":151,"wikipediaUrl":158,"slug":159,"mentionCount":160},"69886e7b033ff25c8c61b12a","SSO","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FSSO","69886e7b033ff25c8c61b12a-sso",6,{"id":162,"name":163,"type":107,"confidence":164,"wikipediaUrl":109,"slug":165,"mentionCount":166},"6a576af1b15b2ddcc32c6b56","Global data- and AI-related fines",0.85,"6a576af1b15b2ddcc32c6b56-global-data-and-ai-related-fines",1,{"id":168,"name":169,"type":107,"confidence":170,"wikipediaUrl":171,"slug":172,"mentionCount":166},"6a576af1b15b2ddcc32c6b55","Model and content liability",0.9,"https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FModel_release","6a576af1b15b2ddcc32c6b55-model-and-content-liability",{"id":174,"name":175,"type":176,"confidence":108,"wikipediaUrl":177,"slug":178,"mentionCount":179},"693feb44312dc892c4c19011","Gartner","organization","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FGartner","693feb44312dc892c4c19011-gartner",115,{"id":181,"name":182,"type":176,"confidence":151,"wikipediaUrl":109,"slug":183,"mentionCount":184},"694e2fe119d266277e1497dc","insurers","694e2fe119d266277e1497dc-insurers",39,{"id":186,"name":187,"type":176,"confidence":151,"wikipediaUrl":109,"slug":188,"mentionCount":189},"6a15d3b3a2d594d36d231641","The Geneva Association","6a15d3b3a2d594d36d231641-the-geneva-association",2,[191,198,205,212],{"id":192,"title":193,"slug":194,"excerpt":195,"category":11,"featuredImage":196,"publishedAt":197},"6a58313a5a245dc50f2b6e4e","Thinking Machines’ Inkling: How a New Open-Weight AI Challenges the Status Quo","thinking-machines-inkling-how-a-new-open-weight-ai-challenges-the-status-quo","Inkling is Thinking Machines’ first general-purpose open-weight model, launched as Western developers seek options beyond closed APIs and Chinese open-weight giants.[1][2] For technical leaders, the r...","https:\u002F\u002Fimages.unsplash.com\u002Fphoto-1618939291225-8d558ea4369f?ixid=M3w4OTczNDl8MHwxfHNlYXJjaHwxfHx0aGlua2luZyUyMG1hY2hpbmVzJTIwb3BlbiUyMHdlaWdodHxlbnwxfDB8fHwxNzg0MTY0NjY2fDA&ixlib=rb-4.1.0&w=1200&h=630&fit=crop&crop=entropy&auto=format,compress&q=60","2026-07-16T01:26:01.779Z",{"id":199,"title":200,"slug":201,"excerpt":202,"category":11,"featuredImage":203,"publishedAt":204},"6a57c28b5a245dc50f2b520b","OpenAI’s Screen‑Free Mobile AI Speaker: What the 2026 Launch Means for the Future of Home Computing","openai-s-screen-free-mobile-ai-speaker-what-the-2026-launch-means-for-the-future-of-home-computing","For years, smart speakers have mostly worked as voice remotes. OpenAI’s planned mobile, screen‑free AI speaker aims to act more like a home computer for the conversational era, built around large lang...","https:\u002F\u002Fimages.unsplash.com\u002Fphoto-1676299081847-824916de030a?ixid=M3w4OTczNDl8MHwxfHNlYXJjaHwxfHxvcGVuYWklMjBtb2JpbGUlMjBzY3JlZW4lMjBmcmVlfGVufDF8MHx8fDE3ODQxMzYzMzF8MA&ixlib=rb-4.1.0&w=1200&h=630&fit=crop&crop=entropy&auto=format,compress&q=60","2026-07-15T17:34:57.211Z",{"id":206,"title":207,"slug":208,"excerpt":209,"category":11,"featuredImage":210,"publishedAt":211},"6a54d65c1d32dce929a1107b","Inside South Korea’s Plan for a Free Nationwide Government AI Chatbot","inside-south-korea-s-plan-for-a-free-nationwide-government-ai-chatbot","South Korea is about to run a national experiment: what happens when a government turns advanced conversational AI into a public utility, as basic as water or electricity?  \n\nUnder the “Everybody’s AI...","https:\u002F\u002Fimages.unsplash.com\u002Fphoto-1619179834700-7a886aac80cc?ixid=M3w4OTczNDl8MHwxfHNlYXJjaHwxfHxzb3V0aCUyMGtvcmVhbiUyMGdvdmVybm1lbnQlMjByZWxlYXNlfGVufDF8MHx8fDE3ODM5NDQ3OTZ8MA&ixlib=rb-4.1.0&w=1200&h=630&fit=crop&crop=entropy&auto=format,compress&q=60","2026-07-13T12:23:09.515Z",{"id":213,"title":214,"slug":215,"excerpt":216,"category":11,"featuredImage":217,"publishedAt":218},"6a4ed95a19d1de4035ab69c8","Inside SpaceXAI and Cursor’s First Joint AI Model Launch","inside-spacexai-and-cursor-s-first-joint-ai-model-launch","1. Context: Why the SpaceXAI–Cursor Model Launch Matters Now\n\nSpaceXAI and Cursor plan to release their first jointly built large language model, Grok 4.5, as early as Wednesday, July 8, after a short...","https:\u002F\u002Fimages.unsplash.com\u002Fphoto-1739036868260-c26b292cd85d?ixid=M3w4OTczNDl8MHwxfHNlYXJjaHwxNnx8YXJ0aWZpY2lhbCUyMGludGVsbGlnZW5jZSUyMHRlY2hub2xvZ3l8ZW58MXwwfHx8MTc4MzU1MjM0Nnww&ixlib=rb-4.1.0&w=1200&h=630&fit=crop&crop=entropy&auto=format,compress&q=60","2026-07-08T23:19:29.298Z",["Island",220],{"key":221,"params":222,"result":224},"ArticleBody_tyGI5hx56eH4zSpYavHDgl70s86f2NaQJfoluT0NPDs",{"props":223},"{\"articleId\":\"6a5768d9b14fe5915b3ecf3e\"}",{"head":225},{}]