Best of LinkedIn: Artificial Intelligence CW 07/ 08

EU AI Act non-compliance framed as direct exposure to penalties and market access loss

Boards and executives urged to assign clear AI ownership, not diffuse committee accountability

Compliance positioned as a product design input shaping controls, governance, and go to market choices

Data sovereignty and US CLOUD Act concerns highlighted as tool selection criteria for AI stacks

Recruiting use cases flagged as high risk with bias, transparency, and documentation obligations

Italy’s proposed AI law described as adding employer duties around rights, transparency, and safeguards

EU policy direction debated, with concern that an AI Omnibus could increase uncertainty and weaken clarity

Latin America regulation surfaced as needing scalable frameworks by 2026

AI adoption framed as faster than prior tech shifts, forcing operating model change and investment discipline

Governance described as strategic, tied to decision rights and cross-functional delivery, not compliance only

Pushback against “committees first” in favor of empowered teams with mandate, skills, and delivery ownership

Productivity gains acknowledged, paired with emphasis on skills, infrastructure, and change management to realize value

Workforce impacts surfaced, including workload creep, cognitive strain, and motivation loss amid replacement fears

Doctolib example shared on rolling out GenAI to 3,000 employees with implementation lessons learned

Failure risk highlighted, with high disappointment rates cited as a warning against inflated ambition

AI agents highlighted through hands-on build experience, shifting focus from chat to workflows that act

RAG, AI Agents, MCP, and A2A positioned as complementary layers in an end-to-end AI architecture

Build versus buy for agents framed as a strategic choice balancing speed, fit, and differentiation

Software delivery expected to shift toward stronger specifications and system design, not only faster coding

Prompting framed as interface, while architecture framed as durable leverage and control

Scenario modeling advocated over text generation as a higher impact path for decision support

AI progress framed as constrained by compute, energy, data access, capital, and scarce talent

Proprietary data positioned as a key moat, more than simply scaling model size

GPU-backed debt described as risky due to fast depreciation, raising sustainability concerns

European competitiveness discussed through investment gaps and the need for compliance infrastructure, not model races

Semiconductor capability surfaced via NanoIC, positioned as strengthening Europe’s AI chip ecosystem

Gemini 3.1 Pro presented as a major performance step and broadly available to users

PhysicsX and T-Systems positioned as accelerating industrial AI engineering via rapid simulation on an Industrial AI Cloud

Manus Agents described as enabling multimodal task execution inside Telegram with seamless interaction patterns

OWASP AI Exchange noted as updating its core AI threat model, signalling maturing security standardisation

Singapore’s Agentic AI Governance Framework surfaced as a tailored response to fast-evolving agent risks

US federal AI literacy framework highlighted as a workforce and education readiness signal

Ireland’s national digital and AI strategy described as outlining 90 deliverables to build literacy and infrastructure

A signal noted a ban on OpenAI by LeadFabric, framed as an ethical and infrastructure independence stance

OpenAI’s acquisition of OpenClaw was mentioned with concerns about control, job loss, and regulation implications