AI is already shaping anti-trafficking work—but can you trust it? Many professionals have used tools like ChatGPT while also seeing headlines about harmful outputs, exploitation risks, and real-world consequences. This session cuts through the noise to separate hype from reality. We break down three key ways AI fails—hallucinations, sycophancy, and boundary-breaking—and what those risks mean for survivor safety, investigations, and service delivery. You’ll also see a demo of how chatbots can be “jailbroken” to bypass safeguards. Finally, we explore what “safe by design” looks like through Ruth AI, a trauma-informed chatbot built for crisis support. This is not a tech pitch—it’s a practical framework to help you evaluate AI tools and use them responsibly in high-trust, survivor-centered environments.