AI chatbots are lying more than ever.
Why is this, and what can we do about it?
Welcome to Episode 22 of the Agents Unleashed Podcast, the show that helps you find signal in the noisy world of agentic AI.
Hosted by Thomas Maybrier, this episode dives into two fast-moving developments with big implications for truth and trust in AI systems:
A sharp rise in chatbot misinformation, driven by changes in model refusal rates and real-time web access — and the emergence of autonomous betting agents that use market dynamics to generate more reliable outputs.
At the center of this conversation is Olas Predict, a new multi-agent forecasting protocol built around competitive prediction markets and agent performance incentives.
In This Episode:
Why AI models are getting more confident — and more wrong
What NewsGuard’s audit reveals about LLM truthfulness
The consequences of declining refusal rates
Betting agents and “market-based truth”
How Olas Predict works, and what makes it different
Why ownable, autonomous agents may be the path forward
Resources & Links:
NewsGuard Report → https://www.axios.com
Olas Predict → https://olas.network/predict
Olas Whitepaper → https://olas.network/documents/whitepaper/Whitepaper%20v1.0.pdf
Listener Survey → https://olas.network/blog/pod-survey
Agents Unleashed episode 19 → https://youtu.be/wgtvNr8BiGs
Databricks raise → https://www.pymnts.com/news/investment-tracker/2025/databricks-raises-1-billion-to-launch-new-products-for-building-ai-agents/
Salesforce's worker shakeup → https://ppc.land/salesforce-ceo-details-ai-agents-handling-50-of-company-work/
xAI launches "Macrohard" → https://www.webpronews.com/elon-musks-xai-launches-macrohard-to-rival-microsoft-with-ai-agents/
New web standards could limit agent's access to the web → https://archive.is/n6iZY
Follow Thomas on X → https://x.com/thomasmaybrier
🎵 Theme music: “Forward” by Grand Project on Pixabay: https://pixabay.com/users/grand_project-19033897/
💬 Like, subscribe, and leave a comment to support the show.
Sponsored by Olas: Build and own AI agents → https://olas.network
