Loading story…
Enterprises are increasingly cautious about adopting large language models (LLMs) for critical business functions, despite the widespread availability and hype surrounding artificial intelligence. Security leaders, such as Sujatha S Iyer from ManageEngine, emphasize that while AI is essential and LLMs have their place, their primary enterprise use cases remain in areas like summarization and content generation rather than core decision-making. In scenarios where AI is used to predict outages or detect fraud, explainability becomes a crucial requirement. Traditional AI models are favored because they can provide clear, actionable reasoning behind their outputs, such as identifying specific spikes in website load or server limitations, which helps leaders respond confidently and quickly. Regulatory pressures, especially in sectors like banking and finance, further drive the need for explainable AI, as compliance teams must understand the rationale behind alerts to ensure effective and lawful responses. The demand for transparency is not just a best practice but a regulatory imperative, particularly for applications like credit scoring and fraud detection. Enterprises are also mindful of the costs associated with deploying advanced AI models, with traditional approaches often being more cost-effective and easier to justify to stakeholders. The complexity and opacity of LLMs raise concerns about overfitting, bias, and the potential for unanticipated outcomes, making them less attractive for high-stakes environments. Compliance requirements are evolving rapidly, and organizations must ensure that their AI systems can withstand regulatory scrutiny, which is more readily achieved with traditional, interpretable models. The need for explainability extends to incident response, where clear AI-driven insights can accelerate triage and remediation. Enterprises are also wary of the resource demands and integration challenges posed by LLMs, preferring solutions that align with existing governance frameworks. The balance between innovation and risk management leads many organizations to stick with tried-and-tested AI approaches. As AI becomes more embedded in business processes, the ability to audit and explain decisions is seen as a competitive advantage. Ultimately, the combination of regulatory, operational, and technical factors ensures that traditional AI remains the preferred choice for many enterprises, especially in regulated industries.
Track how attackers are adapting to this technology.
2 events from the most recent confirmed update back to the earliest known activity.
The reports said enterprises were increasingly using LLMs selectively for tasks such as summarization and content generation while relying on classical ML elsewhere. They also noted experimentation with smaller models, mixture-of-experts approaches, licensed datasets, and explainability frameworks such as LIME and SHAP.
By October 2025, reporting said many enterprises were still choosing traditional machine learning for high-stakes business and security applications because of explainability, compliance, and cost considerations. The coverage highlighted outage prediction, fraud detection, and credit scoring as areas where interpretable outputs remain essential.
Follow how adversaries are adapting to this technology, and where it touches your stack today.
2 references tracked. Mallory keeps watching after this page renders.
Map indicators from this story to your assets and identify affected systems in minutes.
Every observed campaign, victim, and pivot linked to actors named in this story.
Malware, exploits, and IOCs connected to the activity described here.
YARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Get matching new stories delivered to your team as they break — not the next morning.
Ask questions about this story and take action on the answers.
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
Enterprises are reassessing how they deploy **AI models** as security, resilience, and operational control concerns grow around externally hosted **large language models**. One report argues that **small language models (SLMs)** are gaining traction because they can run on-premises, reduce exposure of proprietary data, and better support compliance-sensitive workloads, while still complementing larger models for broader reasoning tasks. The piece frames SLM adoption as a response to board-level concerns over data security, intellectual property protection, and dependence on cloud-hosted AI services. A separate article warns that reliance on **cloud-based LLMs** creates concentrated operational and cyber risk, citing the potential for outages, misconfigurations, software defects, and attacks at a small number of hyperscale providers to disrupt many dependent businesses at once. A third reference discusses **semantic invariance failure** in AI agents, where rewording prompts can produce inconsistent outputs, but it is focused on model robustness research rather than the same enterprise deployment and infrastructure risk discussion. The overall theme is substantive and not fluff because it addresses concrete **security, resilience, and reliability risks** in enterprise AI adoption rather than promoting products or offering generic advice.
Jul 12, 2026
Multiple reports and commentaries highlighted a common concern: **AI risk is increasingly driven by governance, explainability, and system integration failures rather than model performance alone**. The material describes how confident but incorrect AI outputs can become a liability when they affect money, operations, or people and cannot be meaningfully explained, challenged, or traced. It also emphasizes the operational gap between **probabilistic AI outputs** and the **deterministic requirements** of software engineering and enterprise workflows, where errors can propagate through dependencies, validation logic, and connected business systems. The coverage focused on the need for stronger controls across the AI lifecycle, including **build-time, deployment-time, and runtime governance**, as well as clearer accountability for procurement, engineering, and business operators. Key risks cited include prompt injection, model misuse, unsafe responses, hidden integration risk, and automated decision cascades caused by misclassification or incorrect field selection. While the sources vary in format, they consistently argue that effective AI security depends on structured system design, constrained workflows, approved models, secure CI/CD and secrets handling, and governance mechanisms that make AI behavior more predictable and auditable.
Jun 29, 2026
The widespread adoption of large language models (LLMs) in enterprise environments has introduced significant security challenges, particularly due to the tendency to over-rely on their outputs and the normalization of risky behaviors. Experts warn that treating LLMs as reliable and deterministic can lead to systemic vulnerabilities, as these models are inherently probabilistic and can be manipulated through techniques such as indirect prompt injection. This normalization of deviance—where unsafe practices become accepted due to a lack of immediate negative consequences—mirrors historical safety failures in other industries and is exacerbated when vendors make insecure design decisions by default. In addition to technical risks, human factors play a critical role in LLM security. Employees may inadvertently expose sensitive data by pasting it into public LLMs, blindly trust AI-generated outputs, or bypass security policies for convenience, making internal misuse a primary concern. While technical controls such as AI governance and access restrictions are important, organizations must also prioritize security awareness training to address the human side of LLM risk. Building a culture of responsible AI use is essential to mitigate both external threats and internal errors associated with LLM deployment.
Jun 29, 2026