How safe is AI? Toward Trustworthy AI
Computer Science / Digital

How safe is AI? Toward Trustworthy AI

As AI systems grow more pervasive, questions of safety and trust demand urgent, practical answers.
+ Add to watchlist - Remove from watchlist
Start 17:00 o'clock
End 00:00 o'clock

At a glance

Technische Universität Dresden (TUD)
ScaDS.AI Dresden/Leipzig
Andreas-Pfitzmann-Bau
1020
Nöthnitzer Straße 46
01187 Dresden (Dresdner Süden)
Website YouTube

Description

This session presents five research efforts spanning the AI safety landscape: a classifier that identifies AI-generated content across model families (IITGnGPT); toxicity detectors for low-resource Indian languages covering 17 fine-grained harm categories across 12 languages (UnityAI-Guard 1.0 & 2.0); a demonstration of backdoor attacks in YOLO, text classifiers, generative models, and translation systems — revealing a systemic adversarial threat surface; and SangrahaTox, a benchmark dataset for auditing multimodal models for stereotypes, bias, and toxicity. Together, these projects chart a path from detecting what AI produces, to exposing how it can be manipulated, to measuring the biases it silently encodes — offering both diagnostic tools and a broader framework for building trustworthy AI.

Please note that the presentation will be held in English.

Information on the event format

Presentation

Stations

Münchner Platz

  • 3 (tram)

Helmholtzstraße

  • 85 (bus)

Further events of the organizer

SQuAI: Answering Scientific Questions…
Technische Universität Dresden (TUD)
Computer Science / Digital

SQuAI: Answering Scientific Questions Reliably

Every year, millions of new scientific articles are published. With this growing volume of knowledge, it becomes increasingly important to have tools that make it easier to access relevant information. This is exactly where SQuAI comes in: the tool uses artificial intelligence to answer scientific questions by searching relevant research articles and summarizing the most important information in a clear and accessible way.

Start 19:00 o'clock
End 00:00 o'clock
LLMpedia – Materializing an LLM’s…
Technische Universität Dresden (TUD)
Computer Science / Digital

LLMpedia – Materializing an LLM’s Encyclopedic Knowledge at Scale

What does a large language model actually know, and how reliable is that knowledge in long-form text? Benchmarks such as MMLU suggest that modern language models are highly factual, but they only test questions that researchers thought to ask. In the LLMpedia project, researchers generate and evaluate encyclopedia-style articles directly from a model’s parametric memory, making it possible to study what the model knows beyond fixed benchmarks.

Start 19:00 o'clock
End 20:00 o'clock
How safe is AI? Toward Trustworthy AI
Technische Universität Dresden (TUD)
Computer Science / Digital

How safe is AI? Toward Trustworthy AI

As AI systems grow more pervasive, questions of safety and trust demand urgent, practical answers.

Start 17:00 o'clock
End 00:00 o'clock