Tag
This survey examines computational nondeterminism in financial AI systems, covering tabular models, graph networks, and LLM-based workflows, and proposes a layered evaluation framework for auditability.
ORACLE is a new agentic framework for early scam anticipation from streaming app usage trajectories. It uses a self-evolving context manager and on-policy self-distillation to detect scams from partial observations over multiple apps and days.
A practical guide to six SQL patterns for detecting transaction fraud in financial data, including velocity checks, impossible travel detection, and other methods. The author shares real-world examples and tuning advice.
This paper evaluates LLM-based simulators as generators of differentially private synthetic data, using PersonaLedger to assess whether LLMs can faithfully reproduce statistical distributions from DP-protected personas. While achieving promising fraud detection utility (AUC 0.70 at ε=1), the study identifies significant distribution drift caused by systematic LLM biases that override input statistics.
SafetyKit launches AI agents powered by OpenAI's GPT-5, GPT-4.1, and specialized techniques to detect fraud and prohibited activity across text, images, and financial transactions with 95%+ accuracy. The solution enables marketplaces and fintech platforms to automate risk detection, policy enforcement, and content moderation at scale.
Stripe deployed GPT-4 across multiple workflows including support customization, developer documentation assistance, and fraud/bad-actor detection in Discord communities, finding the model outperformed human reviewers in several tasks. The company ran an internal hackathon-style initiative with 100 employees to identify and prototype AI-powered features using OpenAI's latest models.