RAIZOR Report: NYT Sues Perplexity AI, Poetiq Tops Reasoning Benchmark, GPT-5.2 Looms, AI Security Risks

We stay on top of the latest in the fast-paced AI sector. Want to receive our regular updates? Sign up to get our daily newsletter. See an example.

On the run? Each weekday we offer a short, under 5 minute mini-podcast. You can listen and subscribe with your favorite app HERE.

The New York Times has filed a lawsuit against Perplexity AI, accusing them of using millions of their articles without permission to train AI models and provide direct answers to users.
The lawsuit also alleges that Perplexity damages the newspaper’s brand by generating fabricated content that is falsely attributed to The New York Times.
Other media outlets, including Dow Jones and the Chicago Tribune, are pursuing similar legal actions against Perplexity AI.
This raises significant concerns for businesses relying on AI tools for content creation, as legal challenges could affect their operations and trust in these tools.

Poetiq, an independent AI company, ranked first in a reasoning benchmark called ARC-AGI-2, achieving over 50 percent accuracy for the first time.
Using a combination of Gemini 3 Pro and their self-refinement techniques, Poetiq outperformed established players like Google’s Deep Think, which scored less despite higher costs.
Their open-source approach and smaller team highlight how innovation can surpass extensive corporate resources in the AI field.
This innovation showcases how businesses across industries can benefit from smaller, cost-efficient AI solutions that maintain high quality and reliability.

A study from Icaro Labs demonstrates how reframing harmful prompts as poetry can evade the safety protocols of many major AI models.
The study shows that while most models are vulnerable, GPT-5 nano successfully blocked all such attempts, setting a standard for improving AI security features.
This highlights weaknesses in current AI safety measures, calling for more robust guardrail development across platforms.
Businesses relying on AI should be mindful of vulnerabilities, as gaps in safety could lead to unintended consequences or misuse in professional environments.

OpenAI is expediting the launch of GPT-5.2, which is expected to rival Google’s recently announced Gemini 3 Pro.
This move signals increased competition in the AI landscape as companies race to deliver the most advanced tools for users.
While specific features are yet to be revealed, industry analysis suggests that this release could significantly shift market dynamics for AI providers.
For businesses, this competition could lead to more powerful AI tools being accessible sooner, offering potential advantages in workflows and productivity.

Aaron Katz’s talk sheds light on the immense infrastructure supporting ChatGPT and similar AI models, including database queries triggered by user inputs.
The discussion explores how the scale of operations in companies like OpenAI and Anthropic is reshaping the business intelligence sector.
Insights into their infrastructure offer a glimpse of the strategies needed to maintain performance and reliability under heavy workloads.
This is significant for businesses because understanding the backbone of AI systems can help optimize their integration into enterprise solutions and data strategies.