AI 규제 시계 늦춘 유럽의회…CIO, 대응 속도 놓고 딜레마
유럽의회가 3월 26일 EU AI법(EU AI Act)의 일부 시행을 연기하기로 표결하면서 이미 혼란스러운 AI 컴플라이언스 환경에 또 다른 불확실성이 더해졌다. 그러나 애널리스트들은 CIO가 규제가 이미 발효된 상황을 전제로 대응해야 한다고 조언했다. 유럽의회 의원들은 기업의 원활한 제도 이행을 지원하기 위해 유럽 당국이 지침과 표준을 마련할 시간을 확보해야 한다는 이유로, 고위험 인공지능(AI) 시스템 관련 규정의 적용을 연기하는 데 찬성했다. 유럽의회와 유럽연합 집행위원회는 이번 연기에 합의했지만, 최종 시행을 위해서는 또 다른 입법 기관인 유럽연합 이사회의 승인이 필요하다. 유럽의회 의원들이 연기하기로 한 첫 번째 기한이 2026년 8월로 예정됨에 따라 CIO는 중요한 계획상의 과제에 직면하게 됐다. 즉, 약속된 지침 없이 기존 마감 시한 전에 서둘러 변화를 도입할 것인지, 아니면 유럽연합 이사회가 연기를 승인하기를 기대하며 기다릴 것인지 결정해야 하는 상황이다. 연기는 면제가 아니다 애널리스트와 컨설턴트들은 기업 CIO가 기다려서는 안 되며, 규제가 이미 시행된 것처럼 대응해야 한다는 데 사실상 의견을 같이했다. 가트너(Gartner)의 부사장 애널리스트 나데르 헤네인은 “기한 연장이 명확해진 점은 긍정적이다. 이전에는 일정이 계속 변동하는 상황이었다”라며 “다만 최종 결정이 기존 마감 시한에 임박해 내려질 예정이어서 조직들은 결국 당초 계획대로 추진할 수밖에 없다”고 설명했다. 이어 “지난해 11월 EU 디지털 옴니버스 제안의 첫 초안이 발표된 이후, 우리는 고객에게 잠재적인 연기를 AI 시스템 목
Could not retrieve the full article text.
Read on CIO Magazine →Sign in to highlight and annotate this article

Conversation starters
Daily AI Digest
Get the top 5 AI stories delivered to your inbox every morning.
More about
eu ai actkorearesearch
La toma de decisiones en tiempos de IA: el nuevo papel de los CEO
Hace tiempo que la inteligencia artificial dejó de ser una promesa de futuro para convertirse en un motor real de aceleración empresarial. En apenas unos años, ha pasado de ser un conjunto de proyectos piloto impulsados por equipos visionarios, a menudo vistos como “experimentos” dentro de las organizaciones, a consolidarse como una capa transversal capaz de redefinir cómo operan las compañías hoy. Desde el marketing hasta las cadenas de suministro, pasando por finanzas o atención al cliente, la IA está permitiendo reducir tiempos, automatizar tareas complejas y ampliar la capacidad de análisis en empresas de prácticamente todos los sectores. El impacto más visible hoy es la velocidad a la que se pueden gestionar grandes volúmenes de información compleja: procesos que antes requerían seman

South Korea is deploying thousands of ChatGPT-enabled social care robots to help elderly people; over-65s now account for ~20% of the country s 51M people (Financial Times)
Financial Times : South Korea is deploying thousands of ChatGPT-enabled social care robots to help elderly people; over-65s now account for ~20% of the country's 51M people South Korea's strained social care system turns to ChatGPT-enabled devices as population ages NEW
Knowledge Map
Connected Articles — Knowledge Graph
This article is connected to other articles through shared AI topics and tags.
More in Models

Decoding Student Dialogue: A Multi-Dimensional Comparison and Bias Analysis of Large Language Models as Annotation Tools
arXiv:2604.04370v1 Announce Type: new Abstract: Educational dialogue is critical for decoding student learning processes, yet manual annotation remains time-consuming. This study evaluates the efficacy of GPT-5.2 and Gemini-3 using three prompting strategies (few-shot, single-agent, and multi-agent reflection) across diverse subjects, educational levels, and four coding dimensions. Results indicate that while multi-agent prompting achieved the highest accuracy, the results did not reach statistical significance. Accuracy proved highly context-dependent, with significantly higher performance in K-12 datasets compared to university-level data, alongside disciplinary variations within the same educational level. Performance peaked in the affective dimension but remained lowest in the cognitiv

Developing Authentic Simulated Learners for Mathematics Teacher Learning: Insights from Three Approaches with Large Language Models
arXiv:2604.04361v1 Announce Type: new Abstract: Large Language Model (LLM) simulations, where LLMs act as students with varying approaches to learning tasks, can support teachers' noticing of student thinking. However, simulations using zero- or few-shot prompting often yield inauthentic knowledge and language, directing teachers to unrealistic reasoning. We evaluate three approaches (Fine-tuning, Multi-agent, and Direct Preference Optimization; DPO) to improve the authenticity and pedagogical utility of simulated students. All approaches improve cognitive and linguistic authenticity, compared with few-shot prompts. Interviews with elementary mathematics pre-service teachers and researchers (\textit{n} = 8) reveal distinct pedagogical affordances. The fine-tuned model produces realistic, b
b8685
[SYCL] Add Q8_0 reorder optimization (~3x tg speedup on Intel Arc) ( #21527 ) Extend the existing reorder optimization to Q8_0. The reorder separates scale factors from weight data for coalesced memory access -- was implemented for Q4_0/Q4_K/Q6_K but Q8_0 was missing. On Arc Pro B70 (Xe2), Q8_0 tg goes from 4.88 to 15.24 t/s (3.1x) on Qwen3.5-27B. BW utilization: 21% -> 66%. The key fix beyond the kernels: Q8_0 was missing from the type check in ggml_backend_sycl_buffer_init_tensor() that allocates the extra struct carrying the reorder flag -- so the optimization was silently skipped. AI (Claude) was used to assist with root cause investigation and writing the kernel code. All code was human-reviewed and tested on real hardware. Fixes: #21517 macOS/iOS: macOS Apple Silicon (arm64) macOS In




Discussion
Sign in to join the discussion
No comments yet — be the first to share your thoughts!