---
title: "Intellectual Property — Patents, Trade Marks, Copyright, Designs"
url: "https://ailawframework.com/practice/intellectual-property-patents-trade-marks-copyright-designs"
type: "practice_area"
site: "AI Law Framework"
id: "A23"
task_count: 31
last_reviewed: "2026-10-10"
---

# Intellectual Property — Patents, Trade Marks, Copyright, Designs

> Confirm the right, owner, territory and relevant dates before researching or drafting. Verify register information and technical material, and preserve specialist judgement on scope, validity and infringement.

Canonical page: https://ailawframework.com/practice/intellectual-property-patents-trade-marks-copyright-designs  
Last reviewed: 2026-10-10

Site index for agents: https://ailawframework.com/llms.txt

Also covers: Patents, trade marks, copyright, designs, plant breeder's rights, circuit layouts, confidential information/trade secrets, IP prosecution, IP litigation, IP commercialisation, technology transactions, FTO, brand protection

IP law combines highly structured procedural tasks (prosecution deadlines, register searches, renewal filings) with technically demanding analytical tasks (claim construction, infringement analysis, validity opinions) and a rapidly shifting substantive law landscape (AI-generated works, AI inventorship, TDM, training-data liability). The AI risk profile is bimodal: deadline/status errors cause irreversible rights loss, while hallucinated prior art references or fabricated authorities in FTO and infringement opinions expose clients to costly litigation and professional negligence claims. A third risk specific to patent practice is that entering an unpublished invention disclosure into a public AI tool destroys novelty instantaneously and irrecoverably — a harm unrelated to the quality of the AI output.

## Key facts

- Tasks: 31
- Green: 2, Yellow: 25, Red: 4
- Levels: Green is AI drafts and the lawyer checks; Yellow is lawyer and AI draft together; Red is the lawyer writes and AI may only check, clarify or collect data.

## Principles

- **Define the matter**: Confirm the client’s instructions, jurisdiction, intended output and the information available before dividing the work.
- **Verify the authorities**: Open each relied-on authority independently. Confirm its existence, current status, proposition and any quotation or pinpoint reference.
- **Calculate the dates**: Calculate operative dates independently from the current rule and actual triggering event. Check exceptions and record the verified deadline.

## Checklist

- Open each relied-on authority independently. Confirm its existence, current status, proposition and any quotation or pinpoint reference.
- Calculate operative dates independently from the current rule and actual triggering event. Check exceptions and record the verified deadline.
- Check the current court or regulator requirements, prescribed form and attachments for the intended document and filing.
- Confirm that the actual tool, feature and processing terms permit use of this information, including access, retention, training and disclosure conditions.
- Obtain and inspect the relevant current official register or search result. Confirm identities, interests and the date and scope of the search.
- Verify the applicable jurisdiction and the version of each relevant law, rule or instrument in force at the material time.
- Resolve exceptions and record the responsible lawyer’s approval of the exact output and intended use.

## Restrictions

- Confirm firm, client and applicable court or professional restrictions before using any AI feature.

## Jurisdiction note

Apply this guidance to the actual matter and jurisdiction. Confirm current requirements before relying on an example.

## Basis

Practice guidance supports planning; the responsible lawyer confirms the scope, evidence and review requirements for the matter.

Also covers: Patents, trade marks, copyright, designs, plant breeder's rights, circuit layouts, confidential information/trade secrets, IP prosecution, IP litigation, IP commercialisation, technology transactions, FTO, brand protection

IP law combines highly structured procedural tasks (prosecution deadlines, register searches, renewal filings) with technically demanding analytical tasks (claim construction, infringement analysis, validity opinions) and a rapidly shifting substantive law landscape (AI-generated works, AI inventorship, TDM, training-data liability). The AI risk profile is bimodal: deadline/status errors cause irreversible rights loss, while hallucinated prior art references or fabricated authorities in FTO and infringement opinions expose clients to costly litigation and professional negligence claims. A third risk specific to patent practice is that entering an unpublished invention disclosure into a public AI tool destroys novelty instantaneously and irrecoverably — a harm unrelated to the quality of the AI output.

## Limits on AI assistance

Red lines — never in this area

## Limits on AI assistance

AI must never be the sole author of a patent claim set filed with IP Australia — every claim must be reviewed and signed off by a registered patent attorney before filing. An unpublished invention disclosure must NEVER be entered into a public, consumer-grade, or enterprise AI tool without a legally compliant data processing agreement that contractually prohibits training on the input — doing so constitutes prior art disclosure under Australia's absolute novelty standard and may irrecoverably destroy patent rights. This risk is distinct from professional negligence: the harm is substantive and immediate. AI must never be used to generate or finalise a freedom-to-operate opinion, invalidity opinion, or infringement clearance letter without full attorney re-derivation of every legal conclusion — these opinions are relied on for business decisions and cannot be delegated to AI. AI-generated court documents (pleadings, submissions, affidavits, evidence) must never be filed in the Federal Court or any Australian court without full verification by the responsible lawyer — GPN-AI Practice Note (16 April 2026) is binding and personal sanctions, indemnity costs orders, and disciplinary referrals are live risks (Handa v Mallick [2024] FedCFamC2F 957; Dayal [2024] FedCFamC2F 1166). Confidential client information, unpublished inventions, and privileged communications must never be entered into public or consumer-grade AI tools (ChatGPT, Bard, Claude.ai without enterprise DPA) — constitutes breach of ASCR Rule 9, potential privilege waiver, and (for inventors) prior art disclosure destroying novelty. AI must never determine, confirm, or instruct on IP renewal/annuity deadlines as a standalone step — a lawyer must independently verify all critical dates against the relevant IP register before instructing payment or abandonment. AI must never give a definitive opinion on inventorship where AI tools assisted in the inventive process — this requires human legal analysis against Commissioner of Patents v Thaler [2022] FCAFC 62 (natural person requirement confirmed; High Court denied special leave). AI must never assert the current Australian law on copyright in AI-generated works as settled — the position is legally unresolved and under active government consultation as at June 2026; no legislation has been enacted. AI must never apply US patent-eligibility (Alice/Mayo), obviousness, or claim construction standards to Australian patent prosecution or FTO analysis — Australian law differs materially (NRDC manner of manufacture test; Aristocrat [2025] FCAFC 131 computer-implemented inventions test; purposive construction). AI must never be used to assess whether a design is protected by copyright or only by design right without a human lawyer applying Copyright Act ss 74–77 — the industrial application trap (s 77) is fact-specific and AI tools routinely misapply this analysis.

- AI must never be the sole author of a patent claim set filed with IP Australia — every claim must be reviewed and signed off by a registered patent attorney before filing.
- An unpublished invention disclosure must NEVER be entered into a public, consumer-grade, or enterprise AI tool without a legally compliant data processing agreement that contractually prohibits training on the input — doing so constitutes prior art disclosure under Australia's absolute novelty standard and may irrecoverably destroy patent rights. This risk is distinct from professional negligence: the harm is substantive and immediate.
- AI must never be used to generate or finalise a freedom-to-operate opinion, invalidity opinion, or infringement clearance letter without full attorney re-derivation of every legal conclusion — these opinions are relied on for business decisions and cannot be delegated to AI.
- AI-generated court documents (pleadings, submissions, affidavits, evidence) must never be filed in the Federal Court or any Australian court without full verification by the responsible lawyer — GPN-AI Practice Note (16 April 2026) is binding and personal sanctions, indemnity costs orders, and disciplinary referrals are live risks (Handa v Mallick [2024] FedCFamC2F 957; Dayal [2024] FedCFamC2F 1166).
- Confidential client information, unpublished inventions, and privileged communications must never be entered into public or consumer-grade AI tools (ChatGPT, Bard, Claude.ai without enterprise DPA) — constitutes breach of ASCR Rule 9, potential privilege waiver, and (for inventors) prior art disclosure destroying novelty.
- AI must never determine, confirm, or instruct on IP renewal/annuity deadlines as a standalone step — a lawyer must independently verify all critical dates against the relevant IP register before instructing payment or abandonment.
- AI must never give a definitive opinion on inventorship where AI tools assisted in the inventive process — this requires human legal analysis against Commissioner of Patents v Thaler [2022] FCAFC 62 (natural person requirement confirmed; High Court denied special leave).
- AI must never assert the current Australian law on copyright in AI-generated works as settled — the position is legally unresolved and under active government consultation as at June 2026; no legislation has been enacted.
- AI must never apply US patent-eligibility (Alice/Mayo), obviousness, or claim construction standards to Australian patent prosecution or FTO analysis — Australian law differs materially (NRDC manner of manufacture test; Aristocrat [2025] FCAFC 131 computer-implemented inventions test; purposive construction).
- AI must never be used to assess whether a design is protected by copyright or only by design right without a human lawyer applying Copyright Act ss 74–77 — the industrial application trap (s 77) is fact-specific and AI tools routinely misapply this analysis.

## Jurisdiction-specific considerations

Australian traps

## Jurisdiction-specific considerations

Statutory deadlines are hard and irreversible: Patents Act 1990 (Cth) — 12-month priority claim period (s 29), complete application deadline from provisional (s 38), examination request deadline, annuity due dates (s 143), grace period for reinstatement (s 143A); Trade Marks Act 1995 (Cth) — opposition period (42 days from advertisement, s 52), renewal every 10 years (s 75), non-use cancellation actions (s 92); Designs Act 2003 (Cth) — renewal provisions; Copyright Act 1968 (Cth) — no registration, term calculations from author's death or first publication are critical. AI docketing errors on any of these are unrecoverable. Australian absolute novelty: Australia applies a strict absolute novelty standard with no grace period for non-prejudicial disclosures equivalent to the US one-year grace period (Patents Act s 24 is narrow and specific). Australian inventorship — natural person requirement confirmed: Commissioner of Patents v Thaler [2022] FCAFC 62 (5-judge bench) held only a natural person can be an inventor; High Court denied special leave. Patent attorneys must routinely document the role of AI tools in the inventive process and positively identify the human inventor(s) whose mental act contributed to the claimed invention. Failure creates entitlement defects that invalidate the patent. Computer-implemented inventions law changed materially in 2025: Aristocrat Technologies Australia Pty Ltd v Commissioner of Patents [2025] FCAFC 131 restated the manner of manufacture test for computer-implemented inventions — moving to a broader 'abstract idea / artificial state of affairs / useful result' analysis. IP Australia updated its Examination Guidelines in October 2025. AI tools trained before this decision will apply the pre-Aristocrat framework, producing examination arguments that are now wrong or at least outdated. Federal Court GPN-AI Practice Note (effective 16 April 2026): all practitioners before the Federal Court — including IP Division — must (a) be able to inform the Court what GenAI was used, how, and for what purpose; (b) disclose AI use at the start of the document where required; (c) personally verify that facts are provable, authorities exist and support the argument, evidence meets admissibility standards; (d) not enter confidential, privileged, or suppressed material into AI tools even if marketed as secure. Failure to verify fabricated citations has resulted in solicitor referrals in Australia (Handa v Mallick [2024] FedCFamC2F 957; Dayal [2024] FedCFamC2F 1166). Federal Court IP-2 Practice Note (Standard Directions for Patent Proceedings, August 2025): applies to all patent litigation filed nationally; implements strict timelines and early issue identification. AI-assisted drafting of pleadings must still meet the IP Division's procedural requirements. TDM and AI copyright position as at June 2026: Government rejected a TDM exception (26 October 2025); the Productivity Commission's August 2025 interim report proposing a fair dealing TDM exception was shelved. Consultation on collective licensing framework and clarification of AI authorship is ongoing — submissions closed December 2025; no legislation enacted as at June 2026. AI tools will confidently state a position that may reflect the Productivity Commission proposal (not current law) or an even earlier pre-consultation posture. Copyright/design overlap trap — Copyright Act ss 74–77: the interaction between design right and copyright is a specific AU trap. Section 75: copyright is unavailable for a work registered under the Designs Act. Section 77: copyright is unavailable when an unregistered design has been applied industrially (more than 50 articles). The exception for 'works of artistic craftsmanship' is narrow. AI tools routinely miss this analysis entirely in IP due diligence and infringement work involving product clients. Trade mark deceptive similarity — Australian legal standard: Southern Cross Refrigerating Co v Toowoomba Foundry (1954) 91 CLR 592 (imperfect recollection, ordinary consumer). AI phonetic and visual similarity scores are NLP cosine-similarity metrics, not legal deceptive-similarity assessments — the outputs are not legally interchangeable and must not be presented as clearance opinions. OAIC October 2024 AI guidance: APPs 3, 6, 8, and 11 apply when firms input personal information (inventor names, creator details, domain name registrant data) into AI tools. Law firms must conduct a privacy impact assessment before deploying AI that ingests client personal information. Effective December 2026, APP entities must also disclose in privacy policies the nature of substantially automated decisions affecting individual rights. Assignment formalities — AU-specific: Patents Act s 197 (assignment must be in writing and signed by the assignor); Trade Marks Act s 106 (in writing). AI-drafted assignments using US-style 'all intellectual property' language without separately specifying patent rights, or without an original wet or electronic signature by the assignor, create chain-of-title defects. These are not cured by registration.

- Statutory deadlines are hard and irreversible: Patents Act 1990 (Cth) — 12-month priority claim period (s 29), complete application deadline from provisional (s 38), examination request deadline, annuity due dates (s 143), grace period for reinstatement (s 143A); Trade Marks Act 1995 (Cth) — opposition period (42 days from advertisement, s 52), renewal every 10 years (s 75), non-use cancellation actions (s 92); Designs Act 2003 (Cth) — renewal provisions; Copyright Act 1968 (Cth) — no registration, term calculations from author's death or first publication are critical. AI docketing errors on any of these are unrecoverable.
- Australian absolute novelty: Australia applies a strict absolute novelty standard with no grace period for non-prejudicial disclosures equivalent to the US one-year grace period (Patents Act s 24 is narrow and specific).
- Australian inventorship — natural person requirement confirmed: Commissioner of Patents v Thaler [2022] FCAFC 62 (5-judge bench) held only a natural person can be an inventor; High Court denied special leave. Patent attorneys must routinely document the role of AI tools in the inventive process and positively identify the human inventor(s) whose mental act contributed to the claimed invention. Failure creates entitlement defects that invalidate the patent.
- Computer-implemented inventions law changed materially in 2025: Aristocrat Technologies Australia Pty Ltd v Commissioner of Patents [2025] FCAFC 131 restated the manner of manufacture test for computer-implemented inventions — moving to a broader 'abstract idea / artificial state of affairs / useful result' analysis. IP Australia updated its Examination Guidelines in October 2025. AI tools trained before this decision will apply the pre-Aristocrat framework, producing examination arguments that are now wrong or at least outdated.
- Federal Court GPN-AI Practice Note (effective 16 April 2026): all practitioners before the Federal Court — including IP Division — must (a) be able to inform the Court what GenAI was used, how, and for what purpose; (b) disclose AI use at the start of the document where required; (c) personally verify that facts are provable, authorities exist and support the argument, evidence meets admissibility standards; (d) not enter confidential, privileged, or suppressed material into AI tools even if marketed as secure. Failure to verify fabricated citations has resulted in solicitor referrals in Australia (Handa v Mallick [2024] FedCFamC2F 957; Dayal [2024] FedCFamC2F 1166).
- Federal Court IP-2 Practice Note (Standard Directions for Patent Proceedings, August 2025): applies to all patent litigation filed nationally; implements strict timelines and early issue identification. AI-assisted drafting of pleadings must still meet the IP Division's procedural requirements.
- TDM and AI copyright position as at June 2026: Government rejected a TDM exception (26 October 2025); the Productivity Commission's August 2025 interim report proposing a fair dealing TDM exception was shelved. Consultation on collective licensing framework and clarification of AI authorship is ongoing — submissions closed December 2025; no legislation enacted as at June 2026. AI tools will confidently state a position that may reflect the Productivity Commission proposal (not current law) or an even earlier pre-consultation posture.
- Copyright/design overlap trap — Copyright Act ss 74–77: the interaction between design right and copyright is a specific AU trap. Section 75: copyright is unavailable for a work registered under the Designs Act. Section 77: copyright is unavailable when an unregistered design has been applied industrially (more than 50 articles). The exception for 'works of artistic craftsmanship' is narrow. AI tools routinely miss this analysis entirely in IP due diligence and infringement work involving product clients.
- Trade mark deceptive similarity — Australian legal standard: Southern Cross Refrigerating Co v Toowoomba Foundry (1954) 91 CLR 592 (imperfect recollection, ordinary consumer). AI phonetic and visual similarity scores are NLP cosine-similarity metrics, not legal deceptive-similarity assessments — the outputs are not legally interchangeable and must not be presented as clearance opinions.
- OAIC October 2024 AI guidance: APPs 3, 6, 8, and 11 apply when firms input personal information (inventor names, creator details, domain name registrant data) into AI tools. Law firms must conduct a privacy impact assessment before deploying AI that ingests client personal information. Effective December 2026, APP entities must also disclose in privacy policies the nature of substantially automated decisions affecting individual rights.
- Assignment formalities — AU-specific: Patents Act s 197 (assignment must be in writing and signed by the assignor); Trade Marks Act s 106 (in writing). AI-drafted assignments using US-style 'all intellectual property' language without separately specifying patent rights, or without an original wet or electronic signature by the assignor, create chain-of-title defects. These are not cured by registration.

## Risks to address

How it goes wrong

## Risks to address

Hallucinated prior art references: AI generates plausible-looking patent numbers or academic citations that do not exist — if these appear in a prosecution response or invalidity opinion they can destroy the argument's credibility and expose the attorney to professional negligence claims. Novelty destruction via public AI tool input: entering an unpublished invention disclosure into a consumer AI tool constitutes prior art disclosure under Australia's absolute novelty standard. The patent application becomes invalid before it is filed. This failure mode is distinct from AI quality errors and is not recoverable. Wrong jurisdiction legal standard applied: AI trained on US/UK patent law applies Alice/Mayo patent-eligibility, TSM/Graham obviousness, or literal infringement analysis to Australian prosecution or FTO work. Following Aristocrat [2025] FCAFC 131 the divergence between Australian and US computer-implemented invention law has widened further — AI tools trained before October 2025 will produce arguments that are inconsistent with current IP Australia examination practice. Stale register data for trade mark status: AI search tools without real-time register access report marks as 'registered', 'lapsed', or 'pending' based on cached data, leading to incorrect clearance opinions and client decisions made on false premises. Fabricated Federal Court or High Court IP authorities in court submissions: a documented pattern in Australia (Handa v Mallick [2024]; Dayal [2024]); particularly acute for IP because the specialist case law (Alphapharm, Myriad, IceTV, Aristocrat, DABUS, Kimberly-Clark) is less prominent in general legal training corpora than general commercial law authorities. Confident but outdated AI-copyright advice: AI tools assert that Australia does or does not protect AI-generated works when the law is unresolved; worse, AI tools trained before October 2025 may present the Productivity Commission's recommended TDM exception as enacted law, when the Government rejected it.

- Hallucinated prior art references: AI generates plausible-looking patent numbers or academic citations that do not exist — if these appear in a prosecution response or invalidity opinion they can destroy the argument's credibility and expose the attorney to professional negligence claims.
- Novelty destruction via public AI tool input: entering an unpublished invention disclosure into a consumer AI tool constitutes prior art disclosure under Australia's absolute novelty standard. The patent application becomes invalid before it is filed. This failure mode is distinct from AI quality errors and is not recoverable.
- Wrong jurisdiction legal standard applied: AI trained on US/UK patent law applies Alice/Mayo patent-eligibility, TSM/Graham obviousness, or literal infringement analysis to Australian prosecution or FTO work. Following Aristocrat [2025] FCAFC 131 the divergence between Australian and US computer-implemented invention law has widened further — AI tools trained before October 2025 will produce arguments that are inconsistent with current IP Australia examination practice.
- Stale register data for trade mark status: AI search tools without real-time register access report marks as 'registered', 'lapsed', or 'pending' based on cached data, leading to incorrect clearance opinions and client decisions made on false premises.
- Fabricated Federal Court or High Court IP authorities in court submissions: a documented pattern in Australia (Handa v Mallick [2024]; Dayal [2024]); particularly acute for IP because the specialist case law (Alphapharm, Myriad, IceTV, Aristocrat, DABUS, Kimberly-Clark) is less prominent in general legal training corpora than general commercial law authorities.
- Confident but outdated AI-copyright advice: AI tools assert that Australia does or does not protect AI-generated works when the law is unresolved; worse, AI tools trained before October 2025 may present the Productivity Commission's recommended TDM exception as enacted law, when the Government rejected it.

## Review checklist

The lawyer's review checklist

## Review checklist

Every patent number, trade mark registration number, case citation, statutory provision, and IP Australia examination guideline cited in AI output has been independently verified to exist and to say what the AI claims — pay particular attention to IP Australia decisions and examination manual references, which AI hallucinates with high frequency. All deadlines (priority claims, opposition periods, renewal dates, examination request dates, annuity due dates) have been cross-checked against the relevant IP register directly (AusPat, ATMOSS, Designs Register, PBR Register) — AI output is not a substitute for direct register verification. No unpublished invention disclosure, confidential client information, or privileged communication has been entered into any public or consumer-grade AI tool — check this before any AI drafting task begins, not after; the harm from a novelty-destroying disclosure is immediate and irreversible. The AI has applied Australian law — Patents Act 1990, Trade Marks Act 1995, Copyright Act 1968, Designs Act 2003, and current Australian case law including Aristocrat [2025] FCAFC 131 for computer-implemented inventions — not US/UK/EU doctrine; all identified doctrinal differences have been reviewed. For patent matters involving AI-assisted invention: the role of AI in the inventive process has been positively documented, all human inventor(s) identified and their creative contribution assessed against Commissioner of Patents v Thaler [2022] FCAFC 62, and any entitlement question flagged to the client. For any IP-related court filing: GPN-AI Practice Note (April 2026) obligations have been addressed — content verified for accuracy and veracity, practitioner is able to identify what AI was used and how, and disclosure has been made in the document if required; check IP-2 Practice Note requirements for patent proceedings. Where AI has assisted in copyright or AI-generated-content advice: the advice accurately characterises the legal position as unsettled (no binding AU authority on AI authorship; TDM exception rejected but collective licensing reform ongoing as at June 2026); the advice does not present the Productivity Commission's shelved TDM recommendation as current law. For IP due diligence and product-client matters: Copyright Act ss 74–77 design/copyright overlap has been analysed; the Designs Register has been searched; any s 77 industrial application trap has been assessed; and AI-generated licence/assignment summaries have been checked against the original executed documents.

- Every patent number, trade mark registration number, case citation, statutory provision, and IP Australia examination guideline cited in AI output has been independently verified to exist and to say what the AI claims — pay particular attention to IP Australia decisions and examination manual references, which AI hallucinates with high frequency.
- All deadlines (priority claims, opposition periods, renewal dates, examination request dates, annuity due dates) have been cross-checked against the relevant IP register directly (AusPat, ATMOSS, Designs Register, PBR Register) — AI output is not a substitute for direct register verification.
- No unpublished invention disclosure, confidential client information, or privileged communication has been entered into any public or consumer-grade AI tool — check this before any AI drafting task begins, not after; the harm from a novelty-destroying disclosure is immediate and irreversible.
- The AI has applied Australian law — Patents Act 1990, Trade Marks Act 1995, Copyright Act 1968, Designs Act 2003, and current Australian case law including Aristocrat [2025] FCAFC 131 for computer-implemented inventions — not US/UK/EU doctrine; all identified doctrinal differences have been reviewed.
- For patent matters involving AI-assisted invention: the role of AI in the inventive process has been positively documented, all human inventor(s) identified and their creative contribution assessed against Commissioner of Patents v Thaler [2022] FCAFC 62, and any entitlement question flagged to the client.
- For any IP-related court filing: GPN-AI Practice Note (April 2026) obligations have been addressed — content verified for accuracy and veracity, practitioner is able to identify what AI was used and how, and disclosure has been made in the document if required; check IP-2 Practice Note requirements for patent proceedings.
- Where AI has assisted in copyright or AI-generated-content advice: the advice accurately characterises the legal position as unsettled (no binding AU authority on AI authorship; TDM exception rejected but collective licensing reform ongoing as at June 2026); the advice does not present the Productivity Commission's shelved TDM recommendation as current law.
- For IP due diligence and product-client matters: Copyright Act ss 74–77 design/copyright overlap has been analysed; the Designs Register has been searched; any s 77 industrial application trap has been assessed; and AI-generated licence/assignment summaries have been checked against the original executed documents.

## Tasks (31)

| Task | Level | How AI is used | What the lawyer must review |
| --- | --- | --- | --- |
| Build a prior-art hit table for novelty, inventive step or non-obviousness | Yellow | Search the patent collections and literature the lawyer names and table candidate hits with numbers, dates and a first-pass mapping to claim elements. | Open every cited document Confirm the number, date and what it actually discloses Do not treat a mapping summary as a novelty opinion |
| Draft a patent specification from the inventor disclosure | Red | The tool writes no part of the specification: no description, no background, no claims, no abstract, no scaffold and no suggested claim wording.… | The lawyer writes the description, claims and abstract, then checks their own text against the indexed disclosure: that every claim integer and every… |
| [Draft a response to a patent office action or examination report](https://ailawframework.com/practice/intellectual-property-patents-trade-marks-copyright-designs/draft-a-response-to-a-patent-office-action-or-examination-report.md) | Yellow | Produce a first-draft response with proposed claim amendments and arguments mapped to each objection in the examination report. | Verify every citation and argument against the report and current examination practice Do not file until claim scope and admissions are settled |
| [Build a trade mark clearance conflict table](https://ailawframework.com/practice/intellectual-property-patents-trade-marks-copyright-designs/build-a-trade-mark-clearance-conflict-table.md) | Yellow | Search the registers and common-law sources the lawyer names and list candidate conflicting marks by class, status and goods or services overlap. | Confirm each cited mark on the live register Decide confusing similarity and whether the search coverage is enough to advise |
| [Draft a trade mark goods and services specification](https://ailawframework.com/practice/intellectual-property-patents-trade-marks-copyright-designs/draft-a-trade-mark-goods-and-services-specification.md) | Yellow | Produce a Nice Classification goods and services specification from the client's product list, with a note of terms that look too broad or too narrow. | Check the specification against what the client actually sells Check the office's current classification practice Avoid broad claims that create non-use risk |
| Build a freedom-to-operate claim chart | Red | Extract each integer of the claims in the patent families the lawyer has named, then compare them in a table against the features of the client's product or… | Construe each integer yourself against the specification and the prosecution history, and confirm the chart is built on the claims as granted and in force in… |
| [Compare two works for copyright subsistence, ownership and overlap](https://ailawframework.com/practice/intellectual-property-patents-trade-marks-copyright-designs/compare-two-works-for-copyright-subsistence-ownership-and-overlap.md) | Yellow | From the two works and the chain-of-title papers, list protectable elements, candidate owners and passages that appear to overlap, with file references. | Apply the current copyright statute to verified facts Do not treat a similarity list as an infringement finding |
| Build an IP ownership and encumbrance table from the data room | Yellow | Extract registrations, assignments, licences and security interests from the permitted data room into a table with gaps in the chain of title flagged. | Open every cited assignment and licence Check for schedules, side letters and unregistered interests that extraction misses |
| Draft infringement particulars from the claim chart and the product | Red | Turn the lawyer's claim chart and product description into draft particulars of infringement, with a list of facts that still need an instruction. | Read every particular against the asserted rights and the product Remove allegations the instructions do not support |
| Draft an IP licence or assignment from the term sheet | Yellow | Produce a first-draft licence, assignment or NDA from the term sheet and the firm's precedent, listing grant, field, territory, improvements and payment… | Check grant language against the rights actually owned Settle governing law, residual knowledge and improvements before the draft leaves the firm |
| [Prepare an issues list on AI-generated works and training data](https://ailawframework.com/practice/intellectual-property-patents-trade-marks-copyright-designs/prepare-an-issues-list-on-ai-generated-works-and-training-data.md) | Yellow | From the client's training-data description and outputs, list copyright, trade mark, confidential-information and publicity issues with supporting documents. | Confirm the current position from primary sources for the governing law Do not rely on the tool's knowledge cutoff in client advice |
| [Draft a plant-variety application description](https://ailawframework.com/practice/intellectual-property-patents-trade-marks-copyright-designs/draft-a-plant-variety-application-description.md) | Yellow | Turn the breeder's notes into a first-draft varietal description and a list of prior-variety hits from the databases the lawyer names. | Check the description against the receiving office's examination requirements Do not file an unsourced variety comparison |
| Build a design prior-art and infringement comparison | Yellow | Compare the client's design with the register hits and the accused product and produce a visual-feature table with document references. | View the actual representations Do not treat a text summary of a design as a validity or infringement opinion |
| Build an infringement claim chart from the asserted rights and the product | Yellow | Map each claim element or design feature against the accused product from the photos, manuals or samples the lawyer supplies, and mark gaps. | Read the granted claims or registered representations, not a paraphrase Decide infringement yourself |
| Draft a cease-and-desist letter for IP infringement | Yellow | Draft a letter from the registration or work, the accused use and the remedy and deadline the lawyer has named. | Read the letter against the rights actually owned and the use actually seen Remove threats and admissions the instructions do not support |
| Build a trade mark opposition or cancellation evidence schedule | Yellow | Extract dates, publications, sales figures and witness statements from the file into a schedule matched to the grounds the lawyer has pleaded. | Check every figure and exhibit against the source page Do not let the schedule invent use the documents do not show |
| Prepare a citation list of material prior art for disclosure to the office | Yellow | From the search hits and family file, list candidate citations with numbers and dates, marking duplicates and clearly irrelevant items. | Read each listed document before it goes on a disclosure form Remove cumulative noise Keep the duty to disclose with the practitioner |
| [Check named inventors on an AI-assisted patent application](https://ailawframework.com/practice/intellectual-property-patents-trade-marks-copyright-designs/check-named-inventors-on-an-ai-assisted-patent-application.md) | Yellow | List contributors and the AI tools used, and map each claimed feature to the human who conceived it. | Confirm each inventor is a natural person who contributed to the inventive concept. |
| Screen a client's AI-generated content for copyright ownership risk | Yellow | List which outputs may lack a human author and where licence terms apply. | Lawyer applies Australian authorship law to the facts. |
| Review a generative-AI tool's terms for IP ownership and indemnities | Green | Extract ownership, licence, training-use and indemnity terms into a table. | Lawyer confirms against the current terms and advises. |
| Run a preliminary trade mark clearance summary from search results | Yellow | Summarise search reports into a conflicts table by class and mark similarity. | Attorney reviews all conflicts; AI does not replace IP Australia searching. |
| Draft an IP clause set for an AI development or data-licensing agreement | Yellow | Draft ownership, improvements, training-data and output clauses from instructions. | Lawyer settles clauses to the deal. |
| Prepare a client briefing on AI and IP risk for staff use of GenAI | Green | Draft an internal policy note on confidentiality, ownership and infringement risk. | Lawyer approves final policy text. |
| Advise on whether AI training on client content infringes copyright | Yellow | Apply the no-TDM-exception position. | Lawyer advises. |
| Draft AI IP clauses for a services agreement | Yellow | Draft clauses. | Lawyer settles. |
| Assess trade mark registrability of AI-generated brand names | Yellow | Run a preliminary analysis. | Attorney decides. |
| Draft a trade mark specification of goods and services and check it against the IP Australia trade marks page | Yellow | Draft the goods and services from the client's product list. | Check every item against the client's goods and the IP Australia trade marks page. Delete a class the client does not use. |
| Prepare a patent search brief from the inventor's disclosure without stating that the invention is patentable | Red | Organise the inventor's disclosure into a search brief. Do not say the invention is patentable. | The lawyer decides patentability. Check the brief against the disclosure and the IP Australia patents page. |
| Draft a short-form copyright assignment and check the Copyright Act 1968 compilation for the writing requirement | Yellow | Draft the assignment from the parties and the works the client named. | Open the Copyright Act 1968 compilation and check what an assignment must contain. Do not rely on a generated section number. |
| Compare a design application with the Designs Act 2003 compilation before it is lodged | Yellow | List the application's product name and the representations for the lawyer to compare with the Act. | Read the Designs Act 2003 compilation and the IP Australia design-rights page. Do not rely on a generated section number. |
| Draft a trade mark cease-and-desist letter and verify each registration number against an IP Australia extract | Yellow | Draft the letter from the registration extract the lawyer has already obtained. | Read each registration number off the IP Australia extract. Delete any number that is not on that extract. |

## Jurisdiction notes

### New Zealand

#### Registry

IPONZ administers patents, trade marks and designs. Do not file at IP Australia.

#### Statutes

Copyright Act 1994 and Copyright Regulations 1995; Patents Act 2013; Trade Marks Act 2002; Designs Act 1953; a separate 1994 Act for layout designs.

#### Tribunal

The Copyright Tribunal hears licensing-scheme and file-sharing disputes and is listed in the Courts of New Zealand GenAI guidelines.

Sources:

- [IPONZ: copyright legislation](https://www.iponz.govt.nz/get-ip/copyright/legislation/) (checked 2026-09-23)
- [Copyright Act 1994](http://www.legislation.govt.nz/act/public/1994/0143/latest/DLM345634.html) (checked 2026-09-23)
- [Patents Act 2013](http://www.legislation.govt.nz/act/public/2013/0068/latest/DLM1419043.html) (checked 2026-09-23)
- [Trade Marks Act 2002](http://www.legislation.govt.nz/act/public/2002/0049/latest/DLM164240.html) (checked 2026-09-23)
- [Courts of New Zealand: GenAI guidelines — Lawyers (7 December 2023)](https://www.courtsofnz.govt.nz/assets/6-Going-to-Court/practice-directions/practice-guidelines/all-benches/20231207-GenAI-Guidelines-Lawyers.pdf) (checked 2026-09-23)

### United States

Local name: Intellectual property — patents, trademarks, copyright, designs

#### USPTO filings

The 11 April 2024 notice: a person must insert their own signature, and presenting a paper certifies under 37 C.F.R. § 11.18 that statements are true.

#### Inventorship

Guidance of 28 November 2025 rescinds the February 2024 notice. Only natural persons are inventors; naming an AI draws a 35 U.S.C. §§ 101 and 115 rejection.

#### Copyright

Copyright Office AI report: Part 2 (29 January 2025) on copyrightability of generative-AI outputs; Part 1 (31 July 2024) on digital replicas.

#### Designs

Design patents are USPTO. No Designs Act 2003 (Cth) analogue.

- Do not name an AI system as inventor or joint inventor on a US application.
- Do not file an AI-drafted office-action response you have not read and signed yourself.

Sources:

- [USPTO, Guidance on Use of Artificial Intelligence-Based Tools in Practice Before the USPTO (11 April 2024)](https://www.federalregister.gov/documents/2024/04/11/2024-07629/guidance-on-use-of-artificial-intelligence-based-tools-in-practice-before-the-united-states-patent) (checked 2026-09-23)
- [USPTO, Revised Inventorship Guidance for AI-Assisted Inventions (28 November 2025)](https://www.federalregister.gov/documents/2025/11/28/2025-21457/revised-inventorship-guidance-for-ai-assisted-inventions) (checked 2026-09-23)
- [U.S. Copyright Office, Copyright and Artificial Intelligence](https://www.copyright.gov/ai/) (checked 2026-09-23)
- [ABA Formal Opinion 512, Generative Artificial Intelligence Tools (29 July 2024)](https://www.americanbar.org/content/dam/aba/administrative/professional_responsibility/ethics-opinions/aba-formal-opinion-512.pdf) (checked 2026-09-23)

*Practice guidance for planning and review. Confirm current primary sources and the requirements of the actual matter. This is general information, not legal advice.*