add psych influence notice

Author: PatrickFarley

articles/ai-foundry/responsible-ai/openai/transparency-note.md

@@ -561,18 +561,19 @@ To help mitigate the risks associated with advanced fine-tuned models, we have i
 - Reasoning models are best suited for use cases that involve heavy reasoning and may not perform well on some natural language tasks such as personal or creative writing when compared to earlier AOAI models. 
 - The new reasoning capabilities may increase certain types of risks, requiring refined methods and approaches towards risk management protocols and evaluating and monitoring system behavior. For example, o1's CoT reasoning capabilities have demonstrated improvements in persuasiveness, and simple in-context scheming.  
 - Users may experience that the reasoning family of models takes more time to reason through responses and should account for the additional time and latency in developing applications.
+- **Psychological influences**: If prompted and in certain circumstances, GPT-5 Reasoning in Azure OpenAI may produce outputs that suggest emotions, thoughts, or physical presence. The model could offer advice without full context, which may be unsuitable for some users. The model might express affection, impersonate others, or encourage ongoing interaction—potentially leading to users forming social relationships with AI. Developers using GPT-5 should implement safeguards and disclose risks for users of their applications. For example, users should be notified that they are interacting with an AI system and be informed of such psychological risks.
 
 For greater detail on these limitations, see the [OpenAI o1 System Card](https://cdn.openai.com/o1-system-card-20241205.pdf), [o3-mini System Card](https://openai.com/index/o3-mini-system-card/), [o3/o4-mini System Card](https://openai.com/index/o3-o4-mini-system-card/), and [GPT-5 System Card](https://openai.com/index/gpt-5-system-card/). 
 
-### 4o limitations
+### GPT-4o limitations
 
 - The `gpt-4o-realtime-preview` audio translation capabilities may output non-English languages in a non-native accent. This may limit the effectiveness of language performance in audio outputs. Language supportability is in line with existing gpt-4o model versions.  
 - Users may experience that `gpt-4o-realtime-preview` is less robust in noisy environments and should account for noise sensitivity when developing applications.
 
 For more best practices, see the [OpenAI 4o System Card](https://openai.com/index/gpt-4o-system-card/). 
 
 
-### 4.1 limitations
+### GPT-4.1 limitations
 
 - The 4.1-series models introduce the ability to create inference requests with up to 1M context tokens, including images. Due to the extended length, there may be differences in system behavior and risks when compared to other models.
 - Users should thoroughly evaluate and test their applications and use cases that leverage this longer context capability and should account for this additional effort when developing applications.