Blog

AI Without Sovereignty Is Just Outsourced Intelligence

/ , ,

In his opinion piece, Floyd DCosta argues that enterprises adopting AI often gain capability but lack sovereignty—control over how AI models and data are used—creating long-term risks and dependencies on third-party vendors. He emphasizes AI sovereignty as essential, encompassing governance, transparency, data and model control, operational autonomy, and strategic independence, warning that without it, organizations may inadvertently cede their competitive intelligence and face regulatory and operational challenges.

https://www.cio.com/article/4147102/ai-without-sovereignty-is-just-outsourced-intelligence.html

Deterministic AI: What It Is and When to Use It

/ ,

Deterministic AI refers to systems that produce the same output every time they receive the same input, combining AI’s ability to interpret data with deterministic workflows that ensure consistency and control. This hybrid approach uses probabilistic AI models to analyze and classify inputs while embedding their outputs in rule-based automation that executes reliably, making it ideal for enterprise workflows needing predictable, repeatable results. Zapier exemplifies this by orchestrating AI-powered workflows that maintain deterministic execution, blending AI’s flexibility in understanding complexity with automation’s dependability.

https://zapier.com/blog/deterministic-ai/

Focus Areas When Implementing Data Protection by Design and by Default in 2026

/ , , , ,

Data protection by design and by default, a key principle of the EU GDPR, remains inconsistently implemented nearly a decade after its adoption, requiring organizations to consider four main factors—state of the art, cost of implementation, processing context, and risks to individuals—for effective compliance. In 2026, evolving technologies and regulations, especially concerning AI, demand a dynamic, risk-based approach that integrates ongoing assessment and adaptation of technical and organizational measures from the system design stage through deployment to safeguard personal data and uphold individuals' rights.

https://iapp.org/news/a/focus-areas-when-implementing-data-protection-by-design-and-by-default-in-2026

Stop Building Security Goals Around Controls

/ , , ,

Devin Rudnicki, CISO at Fitch Group, emphasizes that security goals should be aligned with business outcomes rather than focused solely on controls, advocating for strategies anchored in corporate objectives, real cyber threats, and industry standards. She highlights three key metrics for security programs—value, risk, and maturity—and stresses the importance of presenting risk in actionable terms for leadership, balancing innovation speed with measured risk, and using automation to free human resources for higher-value work.

https://www.helpnetsecurity.com/2026/03/18/devin-rudnicki-fitch-group-ciso-business-alignment/

CISOs Rethink Their Data Protection Strategies

/ ,

Chief Information Security Officers (CISOs) are rethinking their data protection strategies in response to the rapid expansion of artificial intelligence (AI) use, which magnifies the risks to sensitive data through increased data sharing and exposure. Organizations are enhancing data classification, access management, and monitoring tools, adopting zero-trust frameworks, and frequently updating policies to keep pace with evolving technologies, regulatory requirements, and emerging AI-enabled cyber threats, underscoring the critical need for continuous adaptation in data security programs.

https://www.csoonline.com/article/4143384/cisos-rethink-their-data-protection-strategies.html

Shadow AI Risk: How SaaS Apps Are Quietly Enabling Massive Breaches

/ , , ,

A report from Grip Security reveals that all analyzed companies operate SaaS environments embedded with AI, with a 490% year-over-year increase in public SaaS attacks, 80% involving sensitive data. The article highlights how “shadow AI”—agentic AI within SaaS apps often implemented without IT oversight—enables attackers to use stolen OAuth tokens to cascade breaches across multiple organizations, exemplified by the widespread 2025 Salesloft Drift breach, emphasizing the urgent need for better visibility, continuous governance, and risk-based controls of AI in SaaS to prevent massive cascading cybersecurity incidents.

https://www.securityweek.com/the-shadow-ai-problem-how-saas-apps-are-quietly-enabling-massive-breaches/

Companies Say the Risks of ‘Open’ Artificial Intelligence Models Are Worth It

/ ,

The article reports that many companies are adopting open or partially open AI models despite security and governance concerns, because they offer lower cost, greater customization, and more control than proprietary systems. Firms say smaller, adaptable models are often better suited for business-specific tasks, and most organizations use a mix of open and closed models depending on the use case. The article concludes that while open models introduce risks such as supply-chain vulnerabilities and potential backdoors, companies believe the flexibility and performance benefits make those risks manageable.

https://www.wsj.com/cio-journal/companies-say-the-risks-of-open-artificial-intelligence-models-are-worth-it-0d3ee664

AI Still Doesn’t Work Very Well, Businesses Are Faking It, and a Reckoning Is Coming

/ , ,

Experts from AI advisory firm Codestrap warn that enterprise AI applications often fail to deliver expected benefits due to underlying model limitations and lack of proper metrics to assess AI-generated code quality and business content. They predict a reckoning in 8-9 months as AI misuse leads to failures, lawsuits, pricing pressures, and insurance challenges, urging businesses to adopt clearer strategies, measure true outcomes, and address the hype around AI capabilities.

https://www.theregister.com/2026/03/17/ai_businesses_faking_it_reckoning_coming_codestrap/

Every Layer of Review Makes You 10x Slower

/ , ,

The article argues that each additional layer of review in a process slows progress by a factor of ten, primarily due to waiting time rather than effort, and this bottleneck is not alleviated by AI coding tools. While reviews are necessary to maintain quality and reduce costly mistakes as organizations grow, excessive layers can degrade efficiency and mask root causes of problems, leading to a culture that values checks over genuine quality improvement. The author suggests adopting a Deming-inspired approach emphasizing trust, continuous systemic improvements, and modular small teams that produce high-quality components to reduce reliance on slow review cycles and create a more effective, scalable engineering culture.

https://apenwarr.ca/log/20260316

We Are All AI Philosophers Now

/ , ,

The article emphasizes that AI systems inherently carry the biases and values of their creators through design choices, data, and policy decisions, meaning AI is never truly neutral. It calls on IT leaders to recognize that adopting AI is a governance decision that requires disciplined oversight, transparency, and accountability to manage risks and ensure AI-driven decisions align with organizational and societal values.

https://www.cio.com/article/4145026/we-are-all-ai-philosophers-now.html

Scroll to Top