Blue Team Con 2026: Full Schedule

arrow_back View All Dates

12:30pm CDT

Models and More: using data to inform decision making

Sunday September 13, 2026 12:30pm - 12:55pm CDT

Organizations of all types are working to use data to make better decisions. This includes risk management decisions, such as whether to avoid, mitigate, accept, or transfer a particular risk. But what types of data work best? How do correlation and causation impact your risk analysis? Learn from a cyber insurance pro how they balance the speed of modeling and analytics with the deep experience of domain experts to choose what risks to accept. You will walk away with an understanding of how to effectively use different data sources to support risk management in your organization.

Speakers

Amanda Draeger

Principal Cyber Risk Engineer, Liberty Mutual Insurance

Amanda is a Principal Cyber Risk Engineer at Liberty Mutual Insurance. She is an Army vet, has way too many credentials, and likes yarn.

Sunday September 13, 2026 12:30pm - 12:55pm CDT
Swissôtel Chicago - Track 2

Talk Track 2 (Vevey)

Company Liberty Mutual Insurance
Audience Early Career
Subject Governance / Risk / Compliance
Area Talk
global Y

2:00pm CDT

LLMs - How to Trust When We Can’t Verify

Sunday September 13, 2026 2:00pm - 2:50pm CDT

Swissôtel Chicago - Track 3

LLM internals remain black boxes. We can’t fully trace concepts through LLM layers, and chains of thought offer some degree of visibility but can be unreliable and may become less straightforward as models grow more advanced.

I’ll review the current state (at conference-time) of LLM interpretability and challenges to trustworthiness. Then I’ll discuss how to establish and set boundaries for trust in LLM systems that may be inherently hallucinatory, deceptive, or otherwise untrustworthy.

We’ll need to blend old and new techniques, none of which is sufficient on its own. This may include:

- Independent monitoring tools for LLM chains of thought
- Approaches to evaluate and measure AI output (e.g., like the approach NIST is developing)
- Service-level agreements (SLAs) establishing accuracy requirements for green-lighting various AI use cases within an organization
- Requiring review by multiple humans at critical points in process flows, tailored to the use case
- Early warning indicators, near-miss reviews, and incentives for early-warning and near-miss reporting
- Cross-organizational information-sharing on lessons learned (e.g., via ISACs) to prevent individual firms from duplicating the same mistakes in isolation

The goal is for participants to:
1. Understand the current state of LLM interpretability and challenges to trustworthiness
2. Identify strategies for establishing appropriate boundaries for and degrees of trust in LLM systems
3. Develop a list of techniques to tailor for their organizations' needs

Speakers

Stephanie Losi

Independent consultant

Stephanie is an independent technology risk consultant and writes the Risk Musings newsletter. Previously, she worked as a bank examiner and senior bank examiner at the Federal Reserve Bank of New York, focusing on IT and operational risk assessments of large financial firms. Her... Read More →

Sunday September 13, 2026 2:00pm - 2:50pm CDT
Swissôtel Chicago - Track 3 323 E Wacker Dr, Chicago, IL 60601, USA

Talk Track 3 (Montreux)

Audience Intermediate
Subject Governance / Risk / Compliance
global Y

Blue Team Con 2026

12:30pm CDT

Amanda Draeger

2:00pm CDT

Stephanie Losi

Get help with the event