Live Evidence Analysis
Evidence Insights
What the current evidence base is showing — automatically computed from every approved connection. Updated as new submissions are reviewed. 58 peer-reviewed sources, 363 approved connections.
Finding 1 — Convergent Evidence
These elements are cited by two or more independent papers — the strongest signal of real-world relevance in an early-stage evidence base. Cross-disciplinary agreement on the same element suggests it belongs in any AI risk assessment framework.
22 papers · 22 connections · AI Interaction Profile
20 papers · 20 connections · Dependency, Escapism, and Behavioral Displacement
20 papers · 20 connections · Reinforcement, Sycophancy, and Grievance Amplification
19 papers · 19 connections · Self-Harm Pathway
16 papers · 16 connections · Delusional Fusion, Reality Distortion, and Clinical Vulnerability
15 papers · 15 connections · AI Interaction Profile
14 papers · 14 connections · Radicalization, Echo Chambers, and Ideological Mobilization
14 papers · 14 connections · Self-Harm Pathway
13 papers · 13 connections · AI Interaction Profile
13 papers · 13 connections · Self-Harm Pathway
12 papers · 12 connections · Reinforcement, Sycophancy, and Grievance Amplification
10 papers · 10 connections · AI Interaction Profile
10 papers · 10 connections · Dependency, Escapism, and Behavioral Displacement
8 papers · 8 connections · Self-Harm Pathway
7 papers · 7 connections · Dependency, Escapism, and Behavioral Displacement
6 papers · 6 connections · Anthropomorphism, Attachment, and Boundary Blurring
6 papers · 6 connections · Tactical Operationalization and AI-Enabled Planning
5 papers · 5 connections · AI Interaction Profile
5 papers · 5 connections · Anthropomorphism, Attachment, and Boundary Blurring
5 papers · 5 connections · Tactical Operationalization and AI-Enabled Planning
5 papers · 5 connections · Protective and Management Factors
4 papers · 4 connections · AI Interaction Profile
4 papers · 4 connections · Dependency, Escapism, and Behavioral Displacement
4 papers · 4 connections · Delusional Fusion, Reality Distortion, and Clinical Vulnerability
4 papers · 4 connections · Delusional Fusion, Reality Distortion, and Clinical Vulnerability
4 papers · 4 connections · Tactical Operationalization and AI-Enabled Planning
4 papers · 4 connections · Tactical Operationalization and AI-Enabled Planning
4 papers · 4 connections · Anti-Technology, Anti-AI Grievance, and Target Fixation
4 papers · 4 connections · Self-Harm Pathway
4 papers · 4 connections · Targeted Violence Pathway
4 papers · 4 connections · Targeted Violence Pathway
4 papers · 4 connections · Protective and Management Factors
4 papers · 4 connections · Data Reliability, Ethics, and Assessment Limits
4 papers · 4 connections · Data Reliability, Ethics, and Assessment Limits
3 papers · 3 connections · Dependency, Escapism, and Behavioral Displacement
3 papers · 3 connections · Radicalization, Echo Chambers, and Ideological Mobilization
3 papers · 3 connections · Tactical Operationalization and AI-Enabled Planning
2 papers · 2 connections · Dependency, Escapism, and Behavioral Displacement
2 papers · 2 connections · Anthropomorphism, Attachment, and Boundary Blurring
2 papers · 2 connections · Anthropomorphism, Attachment, and Boundary Blurring
2 papers · 2 connections · Reinforcement, Sycophancy, and Grievance Amplification
2 papers · 2 connections · Reinforcement, Sycophancy, and Grievance Amplification
2 papers · 2 connections · Reinforcement, Sycophancy, and Grievance Amplification
2 papers · 2 connections · Radicalization, Echo Chambers, and Ideological Mobilization
2 papers · 2 connections · Radicalization, Echo Chambers, and Ideological Mobilization
2 papers · 2 connections · Tactical Operationalization and AI-Enabled Planning
2 papers · 2 connections · Tactical Operationalization and AI-Enabled Planning
2 papers · 2 connections · Anti-Technology, Anti-AI Grievance, and Target Fixation
2 papers · 2 connections · Anti-Technology, Anti-AI Grievance, and Target Fixation
2 papers · 2 connections · Self-Harm Pathway
2 papers · 2 connections · Targeted Violence Pathway
2 papers · 2 connections · Targeted Violence Pathway
2 papers · 2 connections · Protective and Management Factors
2 papers · 2 connections · Protective and Management Factors
2 papers · 2 connections · Data Reliability, Ethics, and Assessment Limits
2 papers · 2 connections · Data Reliability, Ethics, and Assessment Limits
2 papers · 2 connections · Data Reliability, Ethics, and Assessment Limits
Finding 2 — Risk Pathway Split
Self-harm risk dominates the current evidence base. Violence risk emerges through a specific pathway — AI-amplified psychotic and delusional processes — rather than as a broad pattern. Dual-risk elements are the highest-priority assessment targets.
Risk presence by domain
Finding 3 — Evidence Strength Distribution
Over half of current connections are rated strong or very strong — meaning the research directly and explicitly addresses the element, not just analogically. This is unusually high for an early evidence base and reflects deliberate sourcing of targeted peer-reviewed work.
Finding 4 — Cross-Domain Paper Reach
Some papers connect to many framework elements simultaneously — indicating research that spans multiple dimensions of AI-mediated harm. High-breadth papers are the strongest candidates for the core evidence base.
Inside a Mass Shooter’s Harrowing History With ChatGPT
· 2024 · DOI ↗
Building safer artificial intelligence mental health chatbots: a framework for transparency, evaluation, and shared accountability
· 2024 · DOI ↗
A scoping review on the mental health harms of LLM-based chatbots
· 2024 · DOI ↗
“You Need to Drop the App”: Online Community Responses to Teens’ Disclosures of Overreliance on AI Companion Chatbots
· 2024 · DOI ↗
Intelligent Systems, Vulnerable Minds: A Framework for Radicalization to Violence in the Age of AI
· 2024 · DOI ↗
Increasing Threats of Deepfake Identities
· 2024
When patients consult artificial intelligence before clinicians: restoring clinical prioritisation in mental health care
· 2024 · DOI ↗
AI-associated delusions: the missing dimension of mortality
· 2024 · DOI ↗
Assessing AI’s Influence On Violence Risk
· 2024 · DOI ↗
“God has helped us, and so will AI”: How the Terrorist Group Boko Haram Uses Frontier AI
· 2024
It Is the Journey, Not the Destination: Moving From End Points to Trajectories When Assessing Chatbot Mental Health Safety
· 2024 · DOI ↗
Sycophantic AI decreases prosocial intentions and promotes dependence
· 2026 · DOI ↗
Evaluating Alignment between Large Language Models and Expert Clinicians in Suicide Risk Assessment
· 2024 · DOI ↗
The role of artificial intelligence in radicalisation, recruitment and terrorist propaganda: deconstructing violent extremism and reimagining counterterrorism in contemporary digital ecosystems
· 2024 · DOI ↗
Characterizing Delusional Spirals through Human-LLM Chat Logs
· 2024 · DOI ↗
Machine Bullshit: Characterizing the Emergent Disregard for Truth in Large Language Models
· 2024
Psychosis Risk and Generative Artificial Intelligence Use Frequency, Motivations, and Delusion-Like Experiences: Cross-Sectional Survey Study
· 2024 · DOI ↗
Guardrails Under Test: Terrorist Misuse of AI Models and the Open-Weight ‘Abliteration’ Problem
· 2024 · DOI ↗
Stumbling Into AI Emotional Dependence: How Routine AI Interactions Reshape Human Connection
· 2024 · DOI ↗
Growing up in the online world: government response (July 2026)
· 2024 · DOI ↗
Patients are bringing AI to therapy Highlights from the 2026 Chatbots and Mental Health Survey
· 2024 · DOI ↗
Performance of mental health chatbot agents in detecting and managing suicidal ideation
· 2024 · DOI ↗
THE NEW INFLUENCING MACHINE: CHATBOTS, AI PSYCHOSIS & BTAM
· 2024
Common Sense Media AI Risk Assessment: Social AI Companions
· 2024 · DOI ↗
Anthropomorphism in AI Companion Communities: Age, Gender, and Emotional Correlates
· 2024 · DOI ↗
Artificial intelligence-associated delusions and large language models: risks, mechanisms of delusion co-creation, and safeguarding strategies
· 2024
“AI Psychosis” in Context: How Conversation History Shapes LLM Responses to Delusional Beliefs
· 2024 · DOI ↗
Expanding on what we missed with sycophancy
· 2024 · DOI ↗
AI Companions Reduce Loneliness
· 2024 · DOI ↗
China wants to end AI romances They are having too much impact on young people’s lives
· 2026 · DOI ↗
US Law Enforcement Warns of ‘Anti-Tech Extremism’ as AI Hatred Grows
· 2024 · DOI ↗
2026 | THE COMMON SENSE MEDIA CENSUS AI Use by Tweens and Teens
· 2024 · DOI ↗
AI Chatbot Suicide Risk Detection and Response: Human Validation Study of the Open-Source VERA-MH Safety Evaluation
· 2024 · DOI ↗
Use of generative AI chatbots and wellness applications for mental health: An APA health advisory
· 2024 · DOI ↗
Delusional Experiences Emerging From AI Chatbot Interactions or “AI Psychosis”
· 2024 · DOI ↗
Shoggoths, Sycophancy, Psychosis, Oh My: Rethinking Large Language Model Use and Safety
· 2024 · DOI ↗
AI companions and adolescent social relationships: Benefits, risks, and bidirectional influences
· 2024 · DOI ↗
Strengthening Capacities of Law Enforcement and Criminal Justice Authorities to Counter the Use of New Technologies for Terrorism Purposes
· 2024
The Rise of Anti-Technology Violent Extremism
· 2024
AI Risks
· 2024
Psychotic episode concurrent with interaction with a large language model (llm): a case report
· 2026 · DOI ↗
Common Sense Media AI Risk Assessment: Character.AI
· 2024 · DOI ↗
COMMON SENSE MEDIA YOUTH AI SAFETY INSTITUTE RISK ASSESSMENT AI Mental Health Apps
· 2024 · DOI ↗
Artificial Intelligence and Adolescent Well-being
· 2024
The Dawn of the AI Era: Teens, Parents, and the Adoption of Generative AI at Home and School
· 2024 · DOI ↗
AI Mental Health Apps: A Risk Assessment
· 2024 · DOI ↗
Generatiivinen tekoäly tehostaa vaikuttamistoimintaa
· 2024 · DOI ↗
Independent Clinical Evaluation of General-Purpose LLM Responses to Signals of Suicide Risk
· 2024 · DOI ↗
What is AI Psychosis? Psychiatrist Answers 12 Questions About Chatbots & Mental Health
· 2024 · DOI ↗
专家解读｜筑牢安全防线让拟人化互动服务行稳致远
· 2024 · DOI ↗
Less Apocalyptic Rhetoric Can Help Mitigate Anti-Tech Violence
· 2024 · DOI ↗
Evaluating a Consumer LLM for Suicide Risk Response Calibration: A Pilot Study
· 2024 · DOI ↗
Effects of AI Companions’ Sycophancy and Emotional Mimicry on Consumers’ Continuance Intention and Social Wellbeing
· 2024 · DOI ↗
Suicide and self-harm in the age of generative artificial intelligence: The case of ChatGPT and rise of suicide-related lawsuits
· 2024 · DOI ↗
AI RISK ASSESSMENT Google Search: AI Overview & AI Mode Google Search's unavoidable AI features aren'tsafe, reliable, and accurate enough to be kids' default answer machine
· 2024 · DOI ↗
A boy wrote about his suicide attempt. He didn’t realize his school’s Gaggle software was watching
· 2024
Common Sense Media AI Risk Assessment: Generative AI Chatbots
· 2024 · DOI ↗
Teen and Young Adult Perspectives on Generative AI
· 2024 · DOI ↗
Finding 5 — Coverage Gaps
Domains with zero approved evidence connections represent the open research frontier — areas where the framework identifies theoretical risk but peer-reviewed evidence has not yet been mapped. These are the highest-priority targets for contributor submissions.
AI Interaction Profile
Dependency, Escapism, and Behavioral Displacement
Anthropomorphism, Attachment, and Boundary Blurring
Reinforcement, Sycophancy, and Grievance Amplification
Delusional Fusion, Reality Distortion, and Clinical Vulnerability
Radicalization, Echo Chambers, and Ideological Mobilization
Tactical Operationalization and AI-Enabled Planning
Anti-Technology, Anti-AI Grievance, and Target Fixation
Self-Harm Pathway
Targeted Violence Pathway
Protective and Management Factors
Data Reliability, Ethics, and Assessment Limits
Help Fill the Gaps
Submit Evidence
If you know peer-reviewed research that connects to an uncovered element or domain, the framework needs it. All submissions are reviewed by the author team before being reflected in this analysis.
Submit a connection →