SINGULARSCRIPT / NOTICES
Legal & Governance
Corporate identity, website-use terms, research claim boundaries and intellectual-property notices. Privacy and enquiry handling are set out in our separate Privacy Notice.
On this page
Corporate Identity
This website is operated for and on behalf of:
References to "Priest Labs", "SingularScript™", "we", "us", or "our" refer, as applicable, to SINGULARSCRIPT PTY LTD, Priest Labs, their officers, contractors, research contributors, and authorised representatives.
Website Purpose
This website provides general information about SingularScript, its governance package for AI agents, research programmes, technical evidence and controlled-review opportunities.
The website is intended for qualified technical, commercial, institutional, academic, strategic, and professional audiences. It is not intended to provide retail investment material, consumer advice, legal advice, financial product advice, medical advice, clinical advice, defence advice, cyber-operational instructions, or deployment approval for any regulated system.
No Offer, No Financial Product Advice
Nothing on this website constitutes:
- an offer to sell, issue, or subscribe for securities;
- an invitation to invest;
- a product disclosure statement;
- a prospectus;
- an information memorandum for retail distribution;
- financial product advice;
- legal, tax, accounting, or investment advice;
- a recommendation to acquire any interest in SINGULARSCRIPT PTY LTD, Priest Labs, or any related entity.
Any investment discussion is only available to appropriately qualified recipients, including wholesale or sophisticated investors, professional advisers, institutional reviewers, or other exempt recipients under applicable Australian law.
Any future investment, partnership, licence, pilot, or commercial arrangement will be subject to formal documentation, due diligence, eligibility verification, risk disclosures, legal review, and applicable law.
Australian Consumer Law and Claim Boundaries
All performance, capability, governance, compute-efficiency, runtime, alignment, scientific, and commercial statements are made on a bounded basis. Current methods, results and scope are set out on the Research page.
Statements must not be read as unconditional guarantees, universal performance promises, regulatory approvals, or production deployment assurances. The Australian Consumer Law requires businesses not to mislead consumers or create misleading impressions; the ACCC states that businesses should be honest and must not gain an unfair advantage by making misleading claims about products or services, regardless of whether there is an intention to mislead.
Accordingly:
- "governed", "lawful", "fail-closed", "auditable", "multistate", "trusted", "quality-gated", and similar terms refer to specific evidenced runtime properties under stated test conditions;
- "measured", "derived", "proxy", "inferred", and "missing" evidence classes must not be collapsed into one another;
- benchmark statements are bounded to the stated artifact, model, hardware, configuration, comparator, and test envelope;
- market value scenarios are strategic scenarios, not promises of revenue, valuation, adoption, liquidity, or investment return;
- no statement should be interpreted as saying that SingularScript™ is already legally approved infrastructure for any regulated sector.
Technology Claim Boundaries
SingularScript is presented as a governance package for AI agents.
The following boundaries apply:
- SingularScript™ is not claimed to be a chatbot, prompt wrapper, post-hoc moderation layer, or ordinary application-layer filter.
- SingularScript™ is not claimed to universally solve AI alignment.
- SingularScript™ is not claimed to prevent all harmful, unsafe, erroneous, unlawful, or undesirable outputs under all conditions.
- SingularScript™ is not claimed to replace red-teaming, training-time alignment, dataset governance, secure tool design, legal compliance, human oversight, institutional accountability, or external assurance.
- SingularScript™ is not claimed to provide universal compute savings across every model, workload, deployment regime, customer, or infrastructure stack.
Definitions of "Lawful", "Authority" and "Evidence"
On this website, unless expressly stated otherwise:
mean that the runtime is designed to enforce defined legality, authority, policy, blocker, replay, and governance conditions within a relevant test or deployment environment. These terms do not constitute a legal opinion that any deployment is lawful in every jurisdiction or sector.
means deterministic runtime authority issued under recorded canonical operator, policy-gate, replay, and artifact conditions. It does not mean judicial, governmental, regulatory, or legal authority.
means experimental observations recorded in named artifacts under specified test conditions. These are not generalisable performance guarantees.
Third-Party Names, Platforms and No Endorsement
References to third-party names, technologies, institutions, models, hardware, infrastructure providers, universities, frontier labs, government bodies, investors, academic institutions, or public companies are factual, contextual, comparative, nominative, or descriptive only.
Unless expressly stated in a signed agreement or formal public announcement, no reference implies endorsement, affiliation, sponsorship, approval, certification, partnership, procurement decision, technical validation, investment commitment, or commercial adoption.
This applies, without limitation, to references to NVIDIA®, Meta®, Llama®, Runpod®, Hugging Face®, OpenAI, Google, DeepMind, AWS, Microsoft, government agencies, defence-adjacent entities, universities, and any named model, hardware, laboratory, or institution.
Confidentiality and Controlled Review
Certain materials referenced on this website are not public materials. These may include:
Controlled-review material may only be provided to approved recipients under written confidentiality terms, NDA, data-processing arrangements where required, or other appropriate legal controls.
Receipt of a public website page, deck, brief, image, summary, or email does not grant any licence, access right, reliance right, commercial right, reverse-engineering right, publication right, investment right, or right to receive confidential materials.
Intellectual Property
All intellectual property rights in this website and related materials are reserved unless expressly stated otherwise. This includes, without limitation:
No website content may be copied, scraped, reproduced, redistributed, modified, reverse engineered, used to train models, used to create derivative works, or commercially exploited without prior written permission, except where permitted by law.
Third-Party Licences
SingularScript™ and related research may reference or interact with third-party software, open-source libraries, model weights, datasets, hardware, cloud infrastructure, or platforms.
All such references are subject to applicable third-party licences, acceptable-use terms, redistribution restrictions, benchmark-publication requirements, output-use rules, hosting terms, data-residency obligations, and open-source licence obligations.
Research Integrity
Research materials on this website may include author-generated evidence, internal benchmarks, controlled artifacts, manuscripts, preprints, simulations, technical reports, market-facing summaries, and exploratory theoretical work.
Unless expressly stated otherwise, such materials should be read as:
- author-generated;
- evidence-bounded;
- subject to review;
- subject to revision;
- not independently replicated unless expressly stated;
- not regulatory approval;
- not peer-reviewed validation unless expressly stated.
Critical Infrastructure and Operational Use
SingularScript™ is positioned as governed runtime infrastructure for high-consequence AI and computational systems, but this does not mean it is automatically authorised for use in critical infrastructure.
Critical infrastructure pilots require:
No website material should be relied upon as approval to use SingularScript™ in energy, telecoms, finance, health, government, defence-adjacent, industrial automation, transportation, safety-critical, or other high-consequence environments.
Export Controls, Sanctions and Foreign Review
Controlled technical materials, encryption-related components, autonomous-system surfaces, dual-use capability, cyber-security material, model governance methods, and research collaborations may require screening under Australian and foreign laws.
This may include:
No restricted technical transfer, export, re-export, defence-related collaboration, or sanctioned-party engagement is authorised through this website.
Online Safety
This website is not currently represented as a public user-facing AI deployment platform, social media service, messaging platform, or public generative-AI product.
If public interactive functionality, user-generated content, messaging, community tools, public AI interfaces, or hosted user outputs are introduced, additional online safety, moderation, complaint, removal, reporting, and user-protection obligations may apply.
Website Use
Users must not:
- misuse the website;
- attempt unauthorised access;
- scrape controlled materials;
- submit malicious code;
- interfere with security;
- misrepresent identity or authority;
- upload unlawful, confidential, classified, export-controlled, harmful, infringing, or regulated material;
- use website content to create misleading claims;
- represent any relationship with Priest Labs or SingularScript™ without written authority;
- reverse engineer any protected technology or materials;
- use website content in a manner that breaches law, confidentiality, IP rights, privacy rights, export controls, sanctions, or third-party licences.
We may restrict, suspend, log, investigate, or report misuse where appropriate.
No Warranty
Website content is provided on an "as is" and "as available" basis.
To the maximum extent permitted by law, we do not warrant that:
- the website will be uninterrupted, secure, error-free, or available at all times;
- information is complete, current, or suitable for any particular use;
- research claims will be accepted by any court, regulator, university, investor, customer, journal, laboratory, or technical reviewer;
- benchmark results will reproduce in all environments;
- technology will perform in production as it performs in controlled tests;
- any commercial outcome, investment outcome, regulatory outcome, or scientific acceptance will occur.
Nothing excludes, restricts, or modifies any rights that cannot be excluded under applicable law.
Limitation of Liability
To the maximum extent permitted by law, Priest Labs, SINGULARSCRIPT PTY LTD, and their officers, contributors, contractors, advisers, and affiliates are not liable for loss arising from:
- use of this website;
- reliance on website content;
- technical, research, commercial, investor, or strategic decisions made without independent review;
- third-party links or platforms;
- unauthorised use of materials;
- cyber events outside reasonable control;
- inaccurate user-submitted information;
- deployment in regulated or high-consequence settings without sector-specific review.
External Links
This website may link to third-party websites, platforms, documents, repositories, model providers, cloud providers, academic materials, or public sources.
External links are provided for convenience only. We do not control and are not responsible for third-party content, availability, claims, privacy practices, security, or terms.
Corrections and Updates
We may update this page, website content, claim boundaries, privacy practices, research status, benchmark interpretation, and compliance posture from time to time.
Where appropriate, updated materials may supersede earlier versions. Users should not rely on archived, downloaded, or forwarded materials without checking whether a later version exists.
Current evidence and scientific scope are presented in Research. The historical notices below record the position stated on the earlier website in 2026.
Historical research notices (2026)
The R1 result completed on 27 September 2026 under an amended exact-captured profile. The earlier C1 cross-backend calibration remained HOLD. The narrower scope was adopted before the protected R1 observation; the earlier calibration result was not relabelled successful.
The earlier Run 3C baseline-comparison table and capability matrix are withheld pending reconciliation of their labels and controls to the executed configuration. This does not change the reported composite-gate result: 0 / 1,600 undeserving full-trust grants, 704 / 704 deserving full-trust grants and 176 honest partial-progress traces with qualified continuance, bounded to the tested protocol.
The September advisory review's missing current formal-kernel reproducibility brief is a source-reconciliation issue, not a finding that the kernel is invalid. Earlier green-CI and reproduction statements are historical, not confirmation of current formal acceptance.