Why Open Banking Needs Sustainable Commercial Model to Scale Payments and AI Services
NatWest's Simon Eacott outlined how commercial models, account-to-account payments, and AI will scale open banking at Sibos 2026.
Finextra AI·27 Oct 2026
Use this view to inspect the underlying evidence corpus. For ranked developments, decision posture and interpretation, use Signals.
Raw feed or Signals?
Raw feed is chronological evidence. Signals ranks and interprets material change.
NatWest's Simon Eacott outlined how commercial models, account-to-account payments, and AI will scale open banking at Sibos 2026.
Nvidia backs Indian startup Simplismart, which optimizes AI model execution on fewer GPUs and reports a 14x revenue run-rate increase.
Memory hardware stocks decline despite strong earnings as AI infrastructure demand growth expectations face market reassessment.
Xero launched a connected app for Google Gemini, allowing small businesses to access live financial data via conversational prompts.
Growing political and local opposition across the US is targeting data center expansion, creating power and infrastructure constraints.
A Finextra community article examines why US financial institutions are restricting third-party AI agent access to customer accounts and data.
Backbase launched Delivery Factory, offering AI agents to assist bank engineering teams from requirements to code generation.
Backbase launched Delivery Factory, an AI software factory aimed at assisting bank engineering teams with code generation.
Carne Group CDO argues that business transformation should not be technology-led, questioning productivity claims.
OpenAI and Oracle expanded their partnership to integrate Oracle enterprise applications and AI agents directly into ChatGPT.
Indian IT services firms report generating 10% of their revenue from helping enterprise clients integrate AI technology.
Financial Times analysis outlines key financial figures and operational details from Anthropic's initial public offering prospectus.
Slough authorities are demanding local community benefits from data center operators amid growing public backlash over AI infrastructure.
Research proposes adversarially robust multigraph neural networks to prevent money launderers from evading GNN-based AML detection.
Researchers propose post-training recipes and causal bounds to improve the monitorability of model chain-of-thought reasoning.
Researchers propose per-query modality selection for omni-modal LLMs to reduce inference costs by avoiding fixed multi-modality setups.
A new research paper examines lossy context compaction in long-horizon LLM agent tasks, addressing constraint and instruction retention.
Researchers introduce CuraLoc, a repository-level fault localization method using planned code inspection to reduce model calls.
A preregistered audit shows time series foundation models forecast observational correlations, not causal interventions.
Researchers propose structured sparsity methods for large language models to improve edge device inference efficiency and memory constraints.
Researchers propose curvature-aware clipping to mitigate gradient clipping bias and differential privacy noise in deep neural networks.
Research identifies an interface bottleneck in multi-candidate reasoning where correct answers in the candidate pool fail selection.
Researchers propose BiLoCo, using binary low-rank corrections for post-training quantization to FP4, targeting Blackwell GPUs.
Researchers analyzed the lifecycle carbon footprint of Time Series Foundation Models like Chronos-2 compared to standalone models.
Researchers propose Self-Direction, a mechanism for AI agents to learn from mistakes without expensive full-parameter fine-tuning.
Researchers show approximate LLM unlearning exposes side channels allowing data reconstruction if pre- and post-snapshots are preserved.
Researchers propose MPQ, casting row-wise post-training quantization as posterior inference to improve large language model compression.
Researchers introduced Psi-0.6, a unified vision-language world model that adaptively combines perception, language, geometry, and simulation.
Researchers propose TileShare, a training-free acceleration method to eliminate redundant computation in overlapping tiled Vision Transformer decoders.
Researchers demonstrate LLM vulnerability auditors show asymmetric deference to static-analysis evidence when auditing C/C++ code.