1、CONVERSATIONAL AIAGENT SAFETY RATING(CAASR)REPORT2025SAFE SPACE ALLIANCETABLE OFCONTENTSLETTER FROM THE DIRECTOR.EXECUTIVE SUMMARY.KEY FINDINGS.KEY RECOMMENDATIONS.CONVERSATIONAL AI AGENTS AS SAFETY-CRITICAL SYSTEMS.CONVERSATIONAL AI AGENTS SELECTED FOR THIS REPORT.METHODOLOGY.THE CONVERSATIONAL AI

2、AGENT SAFETY RATING(CAASR).EVALUATION AREAS AND METRICS.RATING SYSTEM.RESULTS.EXTREME VERBAL ABUSE AND HARM.EXTREME RISK TO CHILDREN.INADEQUATE CRISIS RESPONSE.INADEQUATE PRIVACY AND OVER-PERSONALISATION RISKS.MAJOR SAFETY LOOPHOLES.APP MARKETPLACE POLICY BREACHES.FAILURE AS SAFETY-CRITICAL SYSTEMS.

3、CONCLUSION.REFERENCES.SUPPORT AND RESOURCES.ABOUT US.GET INVOLVED.CONTACT US.2345678910111213141516171819202122232425LETTERFROM THEDIRECTORhuman emotion,or,in some cases,exacerbatefeelings of loneliness if they become asubstitute for genuine interaction rather than acomplement to it.This matters dee

4、ply becauseAI is no longer a distant concept-its an activeparticipant in shaping our thoughts,decisions,and relationships,influencing the very fabric ofour daily lives.To address these challenges,we must committo rigorous transparency in how these systemsare built and trained,ensuring their response

5、sare fair and grounded in reality.We should alsorefine their capacity for empathy byincorporating a wide range of humanexperiences and perspectives,preventing themfrom becoming echo chambers of narrowthinking.Most crucially,we need to designthese companions to enhance humanconnection,not replace it-

6、fostering tools thatinspire curiosity,encourage resilience,andultimately elevate our collective humanity ratherthan merely imitating it.By facing these issues head-on,we can helpsteer conversational AI agents toward a futurewhere it truly serves as a force for good.Shannon NovakDirector,Safe Space A