Claude, by Anthropic, is one of the leading Large Language Models (LLMs). It has been evolving fast, and many updates are relevant for therapy, special education, psychology, and related fields. Here’s a summary of what’s new with Claude, plus ideas (and cautions) for how professionals like us can use it.
Recent updates in Claude
- Claude 4: Opus 4 & Sonnet 4
In mid-2025, Anthropic introduced two upgraded models: Claude Opus 4 and Claude Sonnet 4. These offer better reasoning, improved tool integrations, and more precise following of instructions. Anthropic+2Google Cloud+2 They also support extended thinking (i.e. working with external tools, doing background reasoning), better memory of user context, and capabilities in coding and analysis via their API tools. Anthropic - Larger context windows
Claude Sonnet 4 (for API customers) now can handle a 1 million token context window (Tokens = small chunks of text, like parts of words or short words. 1M tokens = thousands of pages.). That means much longer inputs (texts, reports, previous work) can be fed in as context, which helps with continuity and reduces the need to repeatedly bring the model up to speed. TechCrunch - Memory and project-continuity features
Claude has introduced or upgraded features that allow remembering past conversations, user preferences, uploaded files/projects, etc. Importantly, these are optional (you can disable memory), but for ongoing work (e.g. tracking progress, refining goals over time) this can be very powerful. The Verge - Learning modes & “thinking partner” approach
Claude’s newer educational tools (e.g. Claude for Education) emphasize not simply giving direct answers, but guiding users through reasoning, asking questions, supporting deeper understanding (Socratic style). Venturebeat+1 - Safety and ethics enhancements
There are updates to limit harmful or abusive content interactions, including mechanisms for Claude to end conversations that repeatedly violate safety norms. Also attention to privacy (e.g. incognito or non-memory modes) is more robust. The Verge+1
How these can help SLPs, OTs, Special Educators, Psychologists
Here are some practical ways we might use Claude’s recent capabilities, plus what to be cautious about.
| Function / Task | How Claude can support | Things to watch / best practices |
| Goal / IEP Planning | Use Claude to draft or refine Individualized Education Program (IEP) goals, generate multiple options, suggest evidence-based strategies for goals in speech, fine motor, executive functioning, etc. Because of its improved context memory, Claude can remember student profile details across prompts to help maintain coherence. | Always review drafts carefully; ensure the language matches legal/regulatory standards; verify that suggestions are appropriate for the individual child. Don’t rely on AI for diagnosis. Keep sensitive student info anonymized. |
| Therapy Material Creation | Generate therapy stimuli: e.g. social stories, visual supports, worksheets, scripts for practice, prompts for articulation or language, adapted texts. Longer context window means more ability to build complex lesson sets (e.g. a sequence of sessions) without re-uploading all the materials. | Check for accuracy, cultural appropriateness, developmental level. Avoid overly generic content. Use human insight to adapt. |
| Progress Monitoring & Data Analysis | Claude can help pull together progress reports, analyze data (e.g. logs of student performance or assessment scores), spot trends, suggest modifications in therapy plans. With improved reasoning, it might help suggest when progress is stalled and propose alternative interventions. | Be wary of over-interpreting AI suggestions. Ensure data quality. Maintain human responsibility for decisions. |
| Supporting Learning & Generalization | Use learning modes to help students think through tasks: rather than giving answers, Claude can scaffold reasoning, guide metacognitive strategies, support writing reflections. For older students, help them plan writing or projects with step-by-step reasoning. For psychologists, use it for psycho-educational support (e.g. helping students with ADHD plan tasks, break down executive functioning demands). | Important: always ensure student is learning the process, not “cheating” or bypassing thinking. Monitor for bias or content that seems inappropriate. Confirm information (e.g. if medical or psychological content). |
| Administrative / Documentation Efficiency | Use Claude’s upgraded file tools to create formatted documents, progress notes, therapy plans, meeting summaries, parent-friendly reports. Memory and long context help keep consistent details so you don’t keep repeating basic background. | Even here, you need to review for correctness. Also, check confidentiality and data protection policies. For example, do you have permission to include certain data? Ensure work complies with privacy laws. |
What to be cautious about & ethical considerations
- Privacy and security of student data: always anonymize when possible; understand how Claude stores memory; use incognito/non-memory modes when needed.
- Accuracy: AI outputs can be plausible but not always correct. Double-check especially for diagnostic or intervention content.
- Bias and cultural sensitivity: the model is trained on large datasets that may include biases; ensure materials are appropriate for your students’ cultural, language, and developmental background.
- Dependency: The goal is to use AI as support, not replacement. Our professional judgments, experience, and human insights remain central.
What to try soon
- Experiment with Claude’s memory feature for one student or case (if policy allows) to see whether it helps in continuity of goals/materials.
- Use learning mode when working with students toward reasoning or metacognitive goals, to promote active thinking rather than giving answers.
- Try having Claude generate a therapy plan + visual supports + progress tracking template, then adapt/refine.
- Share with colleagues: what works and what doesn’t; build shared prompts or resources.
References
Anthropic. (2025, May 22). Introducing Claude 4. https://www.anthropic.com/news/claude-4 Anthropic
Anthropic. (2025, August 12). Claude Sonnet 4 model now has a 1 million token context window. TechCrunch. TechCrunch
Anthropic. (2025, August 11). Claude AI memory upgrade & incognito mode. The Verge. The Verge
Anthropic. (n.d.). Claude for Education: Reimagining AI’s Role in K-12 Learning. Eduscape. Eduscape