--- title: "How to Use AI Responsibly (2026 Guide) | Dupple" url: "https://dupple.com/learn/how-to-use-ai-responsibly" description: "How to use AI responsibly: covering bias, privacy, misinformation, environmental impact, and workplace ethics. Guidelines for individuals and teams." published: "2026-02-17" updated: "2026-10-21" llms_txt: "https://dupple.com/llms.txt" --- # How to Use AI Responsibly (2026 Guide) Trusted by 800,000+ Techpresso subscribers · Editorial policy · How we make money Written by Louis Corneloup Founder at Dupple, covering AI tools and strategies for 800K+ readers. How our guides are researched . February 17, 2026 · Updated October 2026 9 min read Using AI responsibly means checking facts before you act on them, keeping sensitive data out of the chat, and saying when AI did a substantial part of the work. On a personal ChatGPT plan, turn off Improve the model for everyone if you do not want new chats used for training. Updated October 2026: the EU AI Act dates now follow the Commission's schedule after the AI Omnibus, the hallucination and energy figures use the OpenAI o3 system card and the IEA report, the US executive order is marked as rescinded, and a section on chatbot training settings was added. A May 2025 [McKinsey survey](https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/tech-forward/insights-on-responsible-ai-from-the-global-ai-trust-maturity-survey) of more than 750 leaders across 38 countries put average responsible-AI maturity at 2.0 out of 4. But individual users (the people actually typing prompts into [ChatGPT](https://chatgpt.com/) and [Claude](https://claude.ai/) every day) often have no framework for how to use AI responsibly. That's a problem, because AI tools have real risks: they can produce false information, reflect societal biases, leak sensitive data, and consume significant energy. Using AI effectively and using it responsibly aren't separate skills; they're the same skill. This guide covers the practical risks of AI use and the specific best practices that protect you, your organization, and the people affected by your AI-assisted decisions. ## Five Risks of Not Using AI Responsibly 1 ### AI Hallucinations and Misinformation AI models generate text that sounds confident and authoritative, even when the information is completely fabricated. Those rates are not a live score for the chatbot on your desk, but they show the size of the problem. [OpenAI](https://openai.com/)'s April 2025 [o3 system card](https://cdn.openai.com/pdf/2221c875-02dc-4789-800b-e7758f3722c1/o3-and-o4-mini-system-card.pdf) recorded a 33% hallucination rate on PersonQA and 51% on SimpleQA, about double o1 (16%) on PersonQA, and [OpenAI's deployment page](https://deploymentsafety.openai.com/o3/hallucinations) says that o3 system is no longer in production. Knowledge workers spend an average of 4.3 hours per week fact-checking AI outputs. And in 2024, 47% of enterprise AI users admitted to making at least one major business decision based on hallucinated content. **Best practices:** - Never publish, send, or act on AI-generated facts without verification - Use AI tools with built-in citations ([Perplexity AI](https://dupple.com/learn/how-to-use-perplexity-ai) cites its sources by default) - For critical decisions, cross-reference AI output with primary sources - Be especially skeptical of specific numbers, dates, quotes, and legal citations, as these are where hallucinations are most common and most dangerous Learning how to identify and mitigate AI errors is a core competency for any professional using these tools. [The AI Academy](https://academy.techpresso.co/?utm_source=dupple_blog&utm_medium=blog&utm_campaign=blog_cta) builds this skill into every lesson with real-world examples of what can go wrong. Visit AI Hallucinations and Misinformation 2 ### Bias in AI Outputs AI models learn from training data that reflects existing societal biases. This means AI can produce outputs that are biased by gender, race, age, geography, and other factors, often in subtle ways that aren't immediately obvious. This matters most when AI is used for decisions that affect people: hiring, lending, performance evaluations, content moderation, and customer service. But it also matters in everyday use. If you ask AI to generate a list of "successful entrepreneurs," the output may skew toward a narrow demographic. If you use AI to write job descriptions, it may use language that discourages certain groups from applying. **Best practices:** - Review AI outputs for demographic and cultural bias before using them - When AI makes recommendations about people (candidates, customers, users), verify that the recommendations don't systematically favor or disadvantage any group - Use diverse prompts and perspectives; ask the AI to consider viewpoints from different backgrounds - For hiring or HR use cases, have multiple reviewers check AI-generated content. See our guide on [ChatGPT for resume writing](https://dupple.com/learn/how-to-use-chatgpt-for-resume) for responsible use in job applications 3 ### Privacy and Data Security When you paste text into an AI chatbot, that data may be used to train future models, stored on third-party servers, or potentially accessed by the AI provider's employees. This is not hypothetical. In 2023, Samsung engineers accidentally leaked proprietary source code by pasting it into ChatGPT. Similar incidents have occurred across industries. **What you should never paste into AI tools:** - Personal data (Social Security numbers, medical records, financial details) - Proprietary source code or trade secrets - Confidential business documents (unreleased financials, M&A details) - Client or customer data protected by NDAs or contracts - Credentials, API keys, or passwords **Best practices:** - Check your AI provider's data policy. ChatGPT offers an opt-out for training data; Claude ([Anthropic](https://www.anthropic.com/)) doesn't train on API inputs; enterprise plans typically offer stronger guarantees - Use anonymization: replace real names, company names, and identifiers with placeholders before pasting sensitive content - For organizations, use enterprise AI plans that offer data processing agreements and compliance certifications - Establish a team-wide policy on what can and cannot be shared with AI tools Verdict The rule that matters most here. Treat anything you paste into a chatbot as potentially stored or used for training, as the Samsung code leak showed. Keep personal data, credentials and client material out, and anonymize or set a team policy. If you use AI at work, our guide on [using ChatGPT for work](https://dupple.com/learn/how-to-use-chatgpt-for-work) includes a section on workplace data safety rules. Visit Privacy and Data Security 4 ### Environmental Impact AI's energy footprint is growing rapidly. The [IEA's Key Questions on Energy and AI report](https://iea.blob.core.windows.net/assets/3179f7f8-01f6-4dd6-bffa-c9f7b73f1dc9/KeyQuestionsonEnergyandAI.pdf) says data-centre electricity use grew 17% in 2025, to 485 TWh, and its central case reaches about 950 TWh by 2030. For individual users, the impact per query is small: a median Gemini Apps text prompt uses about 0.24 Wh of electricity, roughly nine seconds of television, [according to Google](https://cloud.google.com/blog/products/infrastructure/measuring-the-environmental-impact-of-ai-inference). But scale matters. Billions of queries per day add up. **Best practices:** - Be intentional with your AI use; ask well-structured prompts that get the answer in fewer iterations - Choose text over image/video generation when possible, since the IEA report says video generation, reasoning, and agent tasks can use hundreds or thousands of times more energy per query than simple text - Support and advocate for AI providers investing in renewable energy and efficiency improvements - For organizations, factor energy usage into AI tool selection, as not all providers are equally efficient Visit Environmental Impact 5 ### Workplace Ethics and Transparency The thorniest questions about AI responsibility are often social, not technical. Should you tell your manager you used AI to write a report? Should you disclose AI assistance in a job application? What happens when AI automates part of a colleague's role? There are no universal answers, but there are principles: **Transparency builds trust.** In most professional contexts, disclosing AI use is the right move, and increasingly, it's expected. 73% of executives in a 2025 Deloitte survey said they'd want employees to disclose significant AI use in their work. **Credit and authorship matter.** If AI wrote 80% of a deliverable, saying "I wrote this" is misleading. Saying "I drafted this using AI and reviewed/edited it" is honest and, in most workplaces, perfectly acceptable. **Job displacement requires conversation, not avoidance.** If AI is making parts of your team's work redundant, the responsible path is to discuss it openly, not to quietly automate and hope no one notices. ## The Regulatory Picture: What You Need to Know ### EU AI Act The [EU AI Act](https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai) is the world's first major AI regulation. It entered into force on August 1, 2024, and most rules became applicable on August 2, 2026, with high-risk dates later (see the [Commission timeline](https://ai-act-service-desk.ec.europa.eu/en/ai-act/timeline/timeline-implementation-eu-ai-act) after the AI Omnibus). Key provisions: - **Prohibited AI practices** (effective February 2025): Includes social scoring, real-time biometric surveillance, and manipulative AI techniques - **Transparency requirements**: Users must be informed when they're interacting with AI (chatbots, deepfakes, AI-generated content) - **High-risk system rules**: AI used in hiring, education, law enforcement, and healthcare must meet strict requirements for testing, documentation, and human oversight. Annex III uses, including hiring and education, apply from December 2, 2027, and high-risk systems embedded in regulated products from August 2, 2028 - **New ban**: AI systems that generate non-consensual intimate images or child sexual abuse material are prohibited from December 2, 2026 Even if you're not in the EU, the AI Act matters. It sets the global standard, and many companies are adopting its requirements worldwide, similar to how GDPR shaped global privacy practices. ### US Approach The US has taken a sector-specific approach rather than passing broad legislation. [NIST](https://www.nist.gov/artificial-intelligence/executive-order-safe-secure-and-trustworthy-artificial-intelligence) records that the executive order on AI (Executive Order 14110, October 30, 2023) was rescinded on January 20, 2025. Individual states, particularly California and Colorado, have introduced AI-specific legislation. A person using a chatbot in the US has no single federal checklist comparable to the EU AI Act, so the data settings and the checklist on this page are the practical rules. ### What This Means for You - If you're building AI-powered products or features, understand the regulatory requirements in your target markets - If you're using AI for hiring, lending, or other high-stakes decisions, document your process and ensure human oversight - Stay informed. Regulations are changing fast ## Does ChatGPT train on your data? On a personal ChatGPT plan (Free, Plus and Pro), OpenAI may train on your chats unless you turn the setting off. Open Settings, then Data controls, then turn off Improve the model for everyone. New conversations after that are not used to train models, and chats already in your history are not deleted. Temporary Chat is the stricter option for one conversation. ChatGPT Business, Enterprise, Edu, and the API are not used to train models by default. Sources: [OpenAI's data controls](https://help.openai.com/en/articles/7730893-data-controls-in-chatgpt) and [how OpenAI uses content](https://help.openai.com/en/articles/5722486-how-your-data-is-used-to-improve-model-performance). On Claude Free, Pro, and Max, training follows a setting you can change under Settings, then Privacy, then Help improve our AI models, per [Anthropic's consumer training article](https://privacy.claude.com/en/articles/10023580-is-my-data-used-for-model-training). Gemini follows Keep Activity: when it is off, future chats are not used to train models unless you send feedback ([Gemini Apps privacy](https://support.google.com/gemini/answer/13594961)). [Perplexity](https://dupple.com/reviews/perplexity)'s opt-out is the AI data retention toggle under Account settings, then Preferences, and it applies to data collected after you turn it off ([Perplexity's data collection page](https://www.perplexity.ai/help-center/en/articles/11564572-data-collection-at-perplexity)). ## A Checklist for Using AI Responsibly Use this before sharing or acting on any important AI-generated output: **Accuracy check:** - Have I verified all facts, statistics, and claims against primary sources? - Have I checked for any invented citations, fake quotes, or fabricated data? **Bias check:** - Does this output fairly represent diverse perspectives? - Could this content disadvantage any group if used as-is? - If this involves decisions about people, has a human reviewed it for fairness? **Privacy check:** - Did I avoid pasting sensitive, personal, or proprietary data? - Is this output safe to share externally? **Transparency check:** - Would I be comfortable if my audience knew AI helped create this? - Have I disclosed AI use where expected or required? **Proportionality check:** - Is AI the right tool for this task, or am I using it out of convenience for something that needs human judgment? ## Building Habits to Use AI Responsibly Using AI responsibly isn't about avoiding AI. It's about using it with awareness. The most effective AI users are also the most careful, because they understand both what these tools can do and where they fall short. Start with one change: add a fact-checking step to your workflow. Before sending or publishing anything AI-generated, spend 2 minutes verifying the key claims. That single habit prevents the most common and most damaging AI mistakes. For content creation workflows that build in quality checks by default, our guide on [generative AI for content creation](https://dupple.com/learn/generative-ai-for-content-creation) covers responsible production pipelines. The goal isn't to slow down your AI use. It's to make sure the speed doesn't come at the cost of accuracy, fairness, or trust. That's the approach [the AI Academy](https://academy.techpresso.co/?utm_source=dupple_blog&utm_medium=blog&utm_campaign=blog_cta) is built around: teaching you to be both effective and responsible with AI, not one at the expense of the other. ## FAQ **What does it mean to use AI responsibly?** Using AI responsibly means verifying AI-generated information before acting on it, protecting sensitive data from AI tools, checking outputs for bias, being transparent about AI use, and considering the environmental impact of your AI usage. **How often does AI produce incorrect information?** Hallucination rates vary by model and task. OpenAI's April 2025 o3 system card recorded 33% on PersonQA and 51% on SimpleQA, about double o1 on PersonQA, though that o3 system is no longer in production. Always fact-check AI outputs before publishing or making decisions. **What data should I never share with AI tools?** Never paste Social Security numbers, medical records, financial account details, proprietary source code, trade secrets, confidential business documents, client data covered by NDAs, or credentials and API keys into AI chatbots. Use anonymization or placeholders when working with sensitive content. **Do I have to tell people when I use AI?** There is no universal legal requirement yet, but transparency is increasingly expected. The EU AI Act requires disclosure when users interact with AI systems. In most professional settings, disclosing significant AI use builds trust and is considered best practice. **Does ChatGPT train on my data?** On Free, Plus, and Pro, OpenAI may use your content to train models unless you turn off Improve the model for everyone under Settings, then Data controls. Business, Enterprise, Edu, and the API are excluded by default. **What is the EU AI Act and does it affect me?** The EU AI Act is the world's first major AI regulation, with most rules applicable since August 2, 2026 and high-risk rules for hiring and education from December 2, 2027. It bans certain AI practices, requires transparency, and sets strict rules for high-risk AI systems. Even if you are outside the EU, many companies are adopting its standards globally, similar to how GDPR influenced privacy practices worldwide. **Learn to use AI effectively and responsibly with practical, structured guidance.** The AI Academy offers courses, tutorials, and hands-on exercises to help you build responsible AI practices, from bias detection to ethical decision-making frameworks. [Start your free 7-day trial →](https://checkout.dupple.com/checkout/dupple-x-yearly-trial/?utm_source=dupple_blog&utm_medium=blog&utm_campaign=blog_cta) Some offers on this page may be paid placements or contain affiliate links. Related Articles Article ### What Is Email Open Rate Under Apple MPP? Understand reported email open rates under Apple Mail Privacy Protection, their measurement limits and better questions for sponsors. Blog Post ### How to Use ChatGPT for Work (2026 Guide) How to use ChatGPT for work: emails, meeting prep, reports, project planning, and daily workflows. Includes prompts, privacy rules, and plan comparison. Article ### What Is Click-to-Open Rate in Email? Learn how click-to-open rate is calculated, why email privacy affects its denominator and what to ask before comparing reports. Blog Post ### How to Choose a Site for a Sponsored Article Compare publishers for a sponsored article using audience fit, clear disclosures, distribution and a measurable campaign brief. Article ### How to Get Adobe Creative Cloud for Free in 2026 How to get Adobe Creative Cloud for free in 2026: student and teacher offers, school licenses, the 7-day trial, free Adobe apps, and the costs that follow. Article ### How to Get Canva Pro for Free in 2026 Four legitimate ways to get Canva Pro for free in 2026: a Pro trial, Canva Education, Canva for Nonprofits and a Business team trial, with Canva's own terms.