GenAI applications have significant underlying complexities and risks that are not visible on the surface, similar to an iceberg.
GenAI has moved from proof-of-concepts in 2023 to production in 2024, leading to emerging risk management challenges.
It is crucial to ensure the accuracy of your own data and be aware of potential biases in external data used to train LLMs.
The unpredictable, non-deterministic output of LLMs requires implementing input and output guardrails to ensure reliable and safe responses.
Using third-party AI agents introduces trust and security risks, as you often cannot inspect the code or verify their behavior.
Implementing RAG or MCP at scale introduces significant security challenges related to authentication, authorization, and overly permissive access.
Unauthorized use of AI tools by employees, or "shadow AI," creates data leak risks that require clear company policies and education.
GenAI systems must be designed with compliance in mind, providing clear audit trails to meet legal and regulatory requirements like GDPR.
Relying on third-party GenAI APIs creates a dependency that requires a disaster recovery plan to handle outages and prevent cascading failures.
The rapid pace of GenAI adoption will likely lead to significant issues like data leaks and outages, forcing organizations to re-evaluate their speed.
msg
Ismaning, Germany
25:17Scott Hanselman
28:30Cheuk Ho
27:02Mirko Ross
41:06Pascal Reddig, TJ Griffiths, Fabian Schmidt, Oliver Winzenried & Matthias Niehoff & Mirko Ross
28:05Toju Duke
26:25Stephan Gillich
35:37Liran Tal
25:19Chris Wysopal
Jobs that call for the skills explored in this talk.
Netflix, Inc.
Collaboration Betters The World GmbH
Ritwik Infotech
Amazon.com, Inc.
Arlington, United States of America
Agama Solutions
Chandler, United States of America
Amazon.com, Inc.
Dallas, United States of America
