Data Portability And Memory Export Intent

Issue 60 Edition 2026-03-01 5 min read

Not accepted General

Sources: 1 • Confidence: Medium • Updated: 2026-04-12 10:14

Key takeaways

The prompt defines the intended scope of the export to include response-style instructions, personal details, projects/goals, tools and frameworks, preferences/corrections, and any other stored context, requesting verbatim preservation where possible.
The prompt instructs the model to output every stored memory and learned context in a single code block, formatted as "[date saved, if available] - memory content", without summarizing, grouping, or omitting entries.
Anthropic’s "import your memories to Claude" feature at claude.com/import-memory is implemented as a prompt.
The text’s stated purpose is to request a complete export of stored memories because the user is moving to another service.

The prompt defines the intended scope of the export to include response-style instructions, personal details, projects/goals, tools and frameworks, preferences/corrections, and any other stored context, requesting verbatim preservation where possible.
The text’s stated purpose is to request a complete export of stored memories because the user is moving to another service.

The prompt instructs the model to output every stored memory and learned context in a single code block, formatted as "[date saved, if available] - memory content", without summarizing, grouping, or omitting entries.
Anthropic’s "import your memories to Claude" feature at claude.com/import-memory is implemented as a prompt.

Does the target system actually return stored-memory records verbatim when instructed to dump all memories, or does it refuse, redact, or provide partial output?
What is the boundary between "stored memory" and "learned context" in this system, and how is each represented and retrievable?
Are there policy or technical controls that prevent full-memory dumps (e.g., per-item access checks, sensitive-data filtering, or category-based restrictions)?
Is the memory import/export feature consistent across accounts and contexts, or does behavior vary by user state, region, plan, or other conditions?
Is there any documented or observable user-facing mechanism to export memories outside of prompting the model (e.g., a dedicated export tool or endpoint)?

Rising user expectation for full, verbatim memory export suggests data portability could become a competitive requirement for consumer AI assistants and related tooling, affecting product roadmaps and compliance workload.
If memory import and export is implemented via prompt, platforms may face a larger prompt injection and data leakage risk surface, increasing demand for stronger memory access controls and auditing.
Ambiguity between stored memory and learned context indicates potential for user confusion and trust issues, implying that clearer user facing controls and transparency features could become differentiators.

Product updates introducing dedicated memory export and import interfaces beyond prompting, with clear definitions of what is included and how it is formatted.
Documentation or observed behavior showing consistent, controllable memory retrieval with per item access checks, sensitive data filtering, and predictable refusal behavior when appropriate.
Security disclosures or roadmap items addressing prompt based memory dump attempts, including mitigation of prompt injection paths to stored memories.

Consistent refusal or heavy redaction when users request complete memory dumps, indicating that verbatim portability via prompting is not feasible or not allowed.
Evidence that memory handling varies widely by account state, region, or plan, undermining reliability of portability as a product expectation.
Clear policy enforcement that stored memory cannot be exported in full, or that only limited categories are exportable, reducing the scope of the portability use case.