Survivors may use AI chatbots to find information about legal processes, safety planning, available services, housing, financial assistance, or other support. AI can be appealing because it is available immediately and can provide answers in a conversational and accessible way. 

However, AI-generated information can be wrong, incomplete, outdated, or not relevant to a survivor’s specific situation. When a survivor strongly believes information provided by AI, workers may find it challenging to correct misinformation without damaging trust. 

The goal is not to convince survivors that AI is always bad or that workers are always right. Instead, workers can help survivors slow down, verify important information, and make informed decisions while maintaining their autonomy. 

How AI Can Fail Survivors 

1. AI can provide inaccurate or made-up information 

AI chatbots can generate information that sounds right but is incorrect. They may provide inaccurate information about laws, legal processes, court procedures, deadlines, or services. They can also cite cases, laws, or resources that do not exist. They can even push back on your corrections or confirm an incorrect answer. A detailed or convincing response is not proof that the information is accurate. 

Here is an example of AI generating incorrect information. 

2. AI has limited local knowledge 

AI may not know what is happening in a survivor’s community right now. For example, it cannot necessarily know whether a shelter has an available bed today, whether a local program has a waitlist, or whether a service has recently changed its eligibility requirements. 

3. AI comes with privacy risks 

Survivors may enter highly personal information about abuse, children, finances, legal matters, or their location into an AI tool. Depending on the tool and its settings, information shared with an AI may be stored for a long period of time, and companies may have access to chats. This means a survivor chat could be used to train other chatbots, sold to another tech company, or disclosed when a subpoena is received. This can create serious privacy and confidentiality risks for survivors. 

Conversation Prompts 

Here are practical steps you can take when you're supporting a survivor who is using an AI tool that has provided incorrect information. 

1. Validate the survivor’s experience 

A survivor may have turned to AI because they needed information quickly, privately, or when other support was not available. You can acknowledge that the chatbot was helpful without assuming that the information it provided is accurate. 

Instead of: 

“You can't trust ChatGPT.” 

You might say: 

“It sounds like the chatbot was helpful when you needed it. Do you want to look at the information it gave you together?” 

If a survivor strongly believes the AI response is correct, repeatedly telling them it is wrong may make them feel dismissed or less willing to engage. Instead, stay curious and make checking the information something you do with the survivor, not for them. 

2. Verify important information together 

AI can be a useful starting point, but important information should be checked before a survivor relies on it to make a decision. 

If the survivor is comfortable sharing the response, look together at: 

• What question or information was given to the AI 
• What the AI said in response 
• Whether it provided a source 
• Whether you can find the original source 
• Whether the information is current 
• Whether it applies to the survivor’s province or territory 
• Whether it comes from an official or trusted source such as a court, government, legal organization, or service-provider website 
• What could happen if the information is wrong 

The way a question is asked can shape an AI response. AI can also sound confident or agree with assumptions in a question even when the information is incomplete or incorrect. 

Once you have helped verify the important information with the survivor, they can decide what to do next. 

3. Focus on the risk, not the technology 

Rather than deciding whether a particular AI tool is “good” or “bad,” focus on what could happen if the information is wrong or incomplete. 

For example: 

“AI may provide general information about Canadian law, but it may not know the rules that apply in your province, court, or particular situation. Let’s check the information before you rely on it.” 

“A chatbot may give you information about shelters or services, but it cannot know whether a space is available right now. Let’s confirm that directly.” 

“Because this information could affect your legal rights or safety, let’s verify it before you make a decision based on the chatbot’s answer.” 

When Extra Caution is Needed 

AI can help survivors find questions to ask. It should not be the only source they rely on for information that could affect their safety, legal rights, or access to services.  

Encourage survivors to verify AI-generated information before acting when it relates to: 

• Legal rights or responsibilities 
• Court processes, deadlines, or documents 
• Protection orders or other legal remedies 
• Police or criminal justice processes 
• Immigration or legal status 
• Child protection or parenting matters 
• Housing or shelter availability 
• Financial assistance or benefits 
• Personal safety 
• Medical or mental health care 
• Sharing evidence or personal information 
• Any decision where incorrect information could increase risk 

For more information about how to help unrepresented parties use AI more safely, see Rise Women's Legal Centre's AI for Unrepresented Parties: Minimizing Risk. 

Remember: Your role is not to compete with AI for credibility. It is to support the survivor in making informed decisions while maintaining trust, safety, and autonomy. 

Technology-Facilitated Gender-Based Violence (TFGBV) is part of a continuum of violence that can be both online and in-person. If you or someone you know is experiencing TFGBV, you are not alone. You can use ShelterSafe.ca to find a shelter/transition house near you to discuss options and create a safety plan. You don’t need to stay in a shelter to access free, confidential services and support.