Generative AI
AI systems capable of creating new content—text, images, code, or data—based on patterns learned from training.
FAQs
What is the difference between generative AI and traditional AI?
Traditional (discriminative) AI classifies, predicts, or makes decisions based on existing data—a fraud detection model classifies transactions as fraudulent or legitimate; a credit model scores a borrower's default probability. Generative AI creates new data that didn't exist before—it generates text, images, code, or synthetic datasets. Traditional AI recognizes patterns to categorize; generative AI uses patterns to create. In finance, traditional AI has been deployed for years in credit scoring and fraud detection; generative AI is enabling newer applications like automated report writing, conversational interfaces with financial data, and synthetic data generation for model training.
How is generative AI being used in investment research?
Investment research applications of generative AI include: automated summarization of earnings call transcripts, analyst reports, and SEC filings (producing concise highlights with key metrics); generating first drafts of research notes based on financial model outputs and comparable company data; creating alternative data narratives (synthesizing signal from multiple alternative data sources into investment theses); drafting client communications and portfolio commentary; translating between financial formats and languages for global investment teams; and stress-testing investment narratives (generating counterarguments to existing investment theses). Regulatory compliance remains the primary constraint—investment research must comply with financial advice regulations (MiFID II, Reg AC) regardless of whether AI or humans produce it.
What are the cybersecurity risks of generative AI for financial institutions?
Generative AI creates new attack vectors for financial institutions: highly personalized phishing emails generated at scale (no longer limited by poor grammar or generic messaging); AI-generated synthetic media (deepfakes of executives authorizing fraudulent wire transfers or account changes); voice synthesis for vishing (voice phishing) attacks bypassing phone-based authentication; automated social engineering scripts that adapt to target responses; generation of realistic fake financial documents for identity theft and loan fraud; and prompt injection attacks against AI systems embedded in banking workflows (attackers craft inputs that manipulate AI behavior to bypass security controls). Financial institutions are developing AI-specific security controls including deepfake detection, behavioral biometrics, and AI-aware anti-phishing training.
Related Terms
Large Language Model
AI system trained on vast text data to understand and generate human language across many tasks.
Prompt Engineering
Craft of designing and optimizing inputs to AI language models to reliably produce desired outputs.
AI Hallucination
AI model generating confident but factually incorrect or fabricated information not grounded in reality.
Retrieval-Augmented Generation
AI technique grounding language model responses in specific retrieved documents to improve accuracy.