From the Anthropic edition of November 2, 2026
Anthropic confirms it added Claude Code prompt instructions on Opus 5.5 to refuse recreations of specific existing games, citing its IP usage policy
Anthropic (reported by The Verge) · AI model provider; Claude Code users · Global
- Publisher
- The Verge
- Original publication
- Nov 2, 2026
- Source retrieved
- 2026-11-03
- Event date
- 2026-10-30
What happened
The Verge reports Anthropic updated Claude Code's system prompt on Opus 5.5 so it will not recreate a specific existing game. An Anthropic staff member said two sentences were added because using Claude to infringe IP violates the usage policy, that original games, genre games and games from a user's own designs and assets should keep working, and that the company will keep adjusting. Anthropic did not confirm the exact wording circulating on X. The Verge asked whether other Claude system prompts are being changed. XDA, which inspected the Claude Code binary on Oct. 30, found the instruction can be delivered remotely and does not appear to forbid decompilation outright.
Why it matters
Limited direct SLED relevance. Cross-tagged to research because university game-design, digital-preservation and software-archaeology programs use Claude Code for ports and emulation, and vendor prompt changes can refuse legitimate teaching or research work without notice. Why now: Anthropic's confirmation was widely reported Nov. 2; the Nov. 1 edition covered the earlier Engadget report, and this adds XDA's finding of remote delivery.
Evidence and measured results
Anthropic staff statement and independent binary inspection (XDA); no data on refusal frequency.
Limitations and uncertainty
Exact prompt wording unconfirmed; scope beyond Opus 5.5 in Claude Code unclear.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-11-03; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Vendor-side instruction changes can block coursework or research that customers thought was permitted.
- Stakeholders
- Game-design and computer science faculty
- Digital preservation and library staff
- Campus IT AI service owner
- Discovery questions
- Do any courses or labs recreate or port existing games?
- How do faculty report AI tool refusals today?
- Who tracks vendor policy changes?
- Value hypothesis
- Interpretation: a clear change-tracking and false-refusal process reduces disruption from vendor rules.
- Potential engagement
- A short review of affected courses and a false-refusal reporting route.
- Unsupported claims
- That teaching or preservation work will never be refused
- That the exact rule text is known
Pre-sales engineering
Role takeaway
- Fit
- Relevant to Claude Code users in research and teaching; minimal for most agencies.
- Architecture and integration
- Server-delivered instructions apply per model in Claude Code; no client configuration controls them.
- Prerequisites
- Inventory of Claude Code use in courses and labs
- Regression task set
- Constraints
- Scope beyond Opus 5.5 unclear
- Rules may change without notice
- Security
- Not a security change.
- Proof of value
- Run representative coursework tasks on Opus 5.5 and an alternative model
- Test decompilation and emulation tasks directly
- Record refusals with request IDs
Delivery
Role takeaway
- Work
- Log the change, test affected coursework, and set up false-refusal reporting.
- Dependencies
- Faculty
- Campus AI service owner
- Anthropic support
- Ownership
- Campus AI service owner
- Skills and adoption
- Faculty learn which requests may trigger refusals and how to report them.
- Governance checkpoints
- Pre-term review of affected assignments
- Change-log entry for vendor prompt changes
- Acceptance criteria
- Proposal: affected courses identified before next term
- Proposal: every reported false refusal logged with request ID
- Risks
- Silent coursework disruption
- Students switching to unmanaged tools
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Behavior can change through server-delivered instructions without a client update, so regression tests on Claude Code workflows matter.
Governance
Who approves, reviews and stays accountable for outcomes?
Institutions should track vendor prompt and policy changes as change events.
Security and privacy
What data, permissions and controls need testing?
Limited.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Students in game-design courses may hit refusals for assignments that recreate classics for study.
Procurement
What should contracts, pricing and exit terms secure?
Ask for notice of system-prompt changes that affect permitted uses.
Operating model
Which teams own the service once it runs?
Faculty need a route to report false refusals.
Publication history
- 2026-11-02Anthropic · Issue 056 resources
Stable resource ID: verge-claude-code-opus-55-prompt-blocks-game-recreation-20261102