~/talk

I Abuse Your LLM Live with 10 Security Flaws

TalkinFrenchat Sunny Tech 2026 · 2026-07-02..07-03· [security, ai-assisted-development]
speakerdeck:egaetan
slides:none on file

A live demo showing ten concrete security vulnerabilities in applications that use large language models (LLMs). The talk uses an existing app to uncover and exploit code-level LLM security flaws, explains what counts as an LLM security issue, clarifies the development team’s responsibilities, and presents practical secure-development best practices.

AI-written summary · about

  1. 0:00Introduction
  2. 3:06OWASP Top Ten LLM Edition
  3. 7:03Prompt system access flaws
  4. 10:38Prompt injection attacks
  5. 13:20Tool integration risks
  6. 16:10Data leakage issues
  7. 19:15Supply chain risks
  8. 23:19Data poisoning risks
  9. 26:38Hallucinations and biases
  10. 30:13RAG weaknesses
  11. 33:13XSS from LLM outputs
  12. 36:08Resource consumption risks
  13. 37:13GitLab Duo flaw
  14. 38:40Security triangle
  15. 40:15Q&A
~/search
type to search 6,518 conferences, 143,918 speakers and 192,133 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
← →
on a talk: 5 seconds back / forward
0–9
on a talk: jump to 0%–90% of the recording
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.