# Is AI safe in schools? Anthropic’s ‘Claude for Teachers’ shakes up the classroom *ABC News* · https://www.youtube.com/watch?v=QPacQgegVG8 ## Summary Anthropic president and co-founder Daniela Amodei speaks with ABC News' Rebecca Jarvis about the launch of Claude for Teachers, a free tool giving K-12 teachers access to Claude Pro for administrative work and tailored curriculum development. Amodei addresses parent concerns, noting Claude remains restricted to users 18 and older because AI's impact on the developing brain is unknown. She also discusses Constitutional AI and Claude's published constitution, model accuracy, energy efficiency, and advice for graduates entering a changing job market. ## Key points - Anthropic launched Claude for Teachers, giving teachers of kindergarten through age 12 free, tailored access to Claude Pro. - Teachers use the tool for administrative paperwork and for developing curriculum tailored to their state and region's requirements. - Claude.ai remains restricted to users 18 and older because Anthropic says not enough is known about AI's impact on the developing brain. - Anthropic uses Constitutional AI to give the models a system of ethics, making them more honest, helpful, and harmless. - Claude's constitution has been published, and earlier versions drew on documents like the UN's Human Rights declaration. - Amodei says models are not foolproof, and teachers are trained to spot when AI might be entering unknown territory. - With every generation of model, Anthropic has become dramatically more efficient at using chips. - Amodei advises graduates to focus on communication, empathy, and flexibility, saying models predominantly supplement rather than replace human labor. ## Notes ### Product Launch - Anthropic announced **Claude for Teachers**, giving teachers of kids from kindergarten through age 12 free, tailored access to Claude Pro. - It was designed with input from educators to help with administrative work and tailoring lesson plans. - Amodei, whose son is entering kindergarten, says she hopes his school adopts the tool. ### How Teachers Use It - Administrative side: filling out paperwork, forms, and back-end classroom work. - Curriculum side: tailoring lessons to the specific requirements of a teacher's state and region. ### Safety and Age Restrictions - Claude.ai remains for users 18 and above only; students under 18 have no direct access. - Amodei says this is not because AI is necessarily bad for kids, but because not enough is known about AI's impact on the developing brain. ### Mistakes and Accountability - Amodei acknowledges models are not foolproof and can hallucinate. - Teachers are trained to spot signs of the AI entering unknown territory and to partner effectively with the tool. - The tools are not designed to replace thinking; teachers retain deep knowledge of curriculum and students. ### Ethics and Constitutional AI - Constitutional AI is used to align models with human values, making them more honest, helpful, and harmless. - Claude's constitution has been published; earlier versions relied heavily on founding documents including the UN's Human Rights declaration. - The rate of harmful responses has dramatically declined over years, per Amodei. ### Energy and Efficiency - No specific timeline for AI solving energy consumption, but each model generation uses chips dramatically more efficiently. - Anthropic has long been a proponent of doing more with less. ### Advice for Graduates - Anthropic looks for communication, empathy, flexibility, and relating to others. - Research shows models predominantly supplement human labor rather than replace humans, changing how work is done. - Younger people who grew up natively with AI may have an edge due to adaptability and human-to-human skills. ## Transcript [00:01] Welcome back. AI giant Anthropic today announced a new teaching tool called Claude for teachers. The new chatbot was designed with input from educators, and the company says it can help with everything from administrative work to tailoring lesson plans. Our Rebecca Jarvis spoke exclusively with Anthropic president and co-founder Daniela Amodei about the making of the new product and why the company thinks it'll be a game-changer for the education system. Amodei also answers questions about Claude's code of ethics and why AI's energy use might decrease with time. >> Tell us about this new tool and how teachers will be using it. >> Today, we are launching Claude for teachers. And this is a tool that allows [00:41] teachers who are in the classroom with kids from kindergarten through age 12 to have free, tailored access to Claude Pro. And this enables them to be able to really take specific curricula and tailor them to the needs of their specific classroom. >> How have teachers there been using it? >> One is really helping teachers on the administrative side. So, helping them to fill out paperwork and forms and do a lot of the kind of back-end work that's needed just in the day-to-day of running a classroom. The second one is helping teachers develop curriculum. So, it's tailored to the specific requirements of that state and region and their educational curriculum to really partner with teachers to help them develop [01:29] tailored learning tools for their specific classroom. >> You're also a mom. Is this something you would hope to see in your children's school someday? >> I personally feel really, really excited for what it means. My son's going into kindergarten this year. And I'm really hopeful that his school will adopt, you know, Claude for teachers to help tailor curriculum for him and for his classmates. >> There are going to be parents who hear this and feel nervous because parents have watched their kids with smartphones and screentime and social media and they've seen the unwanted unintended consequences that they weren't really prepared for. This is in the hands of teachers now, but what do you say to the [02:11] parents who are nervous that this new technology is going to repeat those same mistakes? >> I completely understand and relate to this worry and I think a key difference here is that Claude.ai remains for 18-year-old users and above only. So, individual students below the age of 18 won't have access to the tool directly and I think this isn't because we necessarily feel, you know, worried that this is, you know, a terrible thing to give to kids, but I think the reality is we just don't know enough about AI's impact on the developing brain for Anthropic to feel comfortable giving access to students directly. >> We know that AI can hallucinate. It can make mistakes. If this chatbot were to make a mistake, if it were to give the [03:05] wrong information or bad advice or reinforce bias, who would be accountable for that? >> These models are not foolproof. They do make mistakes and part of how we have worked with teachers in developing this technology and really listening to their questions and concerns is helping to, you know, spot when an AI might be, you know, entering unknown territory. What are the signs to look for and how can that teacher partner most effectively with the tool to make sure that it's being used well. I think the other, you know, most important kind of quality in terms of how we train the models is Constitutional AI is a tool that we have used to really make the models be as aligned with human values as possible, [03:52] which essentially makes the models have the system of ethics. It makes the models more honest, more helpful, and more harmless. And what we've seen over the course of years is that the models' ability to be, uh, you know, accurate in terms of, you know, responding to questions, the rate of like harmful responses has just dramatically declined. So, we're really feeling very positive about the models' abilities to interact in a variety of different situations, uh, with the teachers to make sure that that the answers that are being provided are are accurate. >> Is there a backstop, though? What resources would a teacher have if they weren't sure about a response? >> So, part of what I think is so powerful [04:36] about these tools is that they're not designed to replace thinking. Teachers have a wealth of information about, uh, the curriculum that they're teaching, about the specific students that they are trying to impart that information to. The way I would look at this is Claude is almost like a great, uh, partner in helping the teachers sort through challenging topics or questions or understand part of the curriculum, um, and how to teach that to their students in a way that works for them. >> You also mentioned earlier the idea of ethics, that ethics are a part of the creation of this. Who sets those ethics? What are the guiding principles and North Stars in that ethics category? >> So, we've actually published what [05:22] Claude's constitution looks like, and, uh, just the many different, uh, ideas and documents and people that have really contributed to creating this, uh, this framework for how we teach Claude this this system of ethics. Some earlier versions of our constitution relied very heavily on a number of different founding documents, uh, ranging from things like the UN and of Human Rights. And so, the idea here is that instead of just sort of telling the model, oh, you know, be nice. What does be nice mean, right? There's a huge body of information that humans have built over the course of uh centuries about how to behave in ways that are fair, that are kind, that are ethical, and that are responsible. And we [06:08] incorporated those into how we actually train and reinforce the models' uh capabilities. >> One of those things that that keeps coming up, it comes up every time I talk about AI, people ask me about the energy consumption, the resource consumption. And while data centers have become more efficient over time because of the chips, there's just more demand for AI, which then increases the energy usage and increases the costs on the communities where they are centered. How far off are we from AI solving that problem? Using less energy, the world finding alternatives to energy because of AI. >> We don't know a specific timeline sort of for the future, but what I would say is with every generation of model, we [06:56] have become dramatically more efficient at how well we use the chips. I think Anthropic has, you know, long been a proponent of doing more with less. >> When we last spoke, I was very struck by your advice for students to study the humanities. What do you say to college graduates who are looking for work right now, and the feedback that they're hearing, we're holding off on hiring because AI might be able to do a lot of these entry-level jobs. >> I really think for young people coming out of college, I would continue to say that the qualities that we most look for at a place like Anthropic are an ability to communicate well, to relate to other people, to have empathy, to be flexible, right? Because these AI models are [07:47] really changing the nature of work. So much of what we have seen from the research that we've done is models are still predominantly supplementing human labor. They're not replacing humans. They're changing how we do our work. But I think in a lot of ways, younger people, especially folks who like grew up natively using and understanding artificial intelligence, I actually believe in some ways can have an edge because they're really adaptable. They're curious, they've prioritized, I think a lot of the time, the human-to-human skills that I think not only are not going to go away, but are going to be ascendant in the workplace in a world of of artificial intelligence. >> Our thanks to Rebecca Jarvis.