# AISOF, full reference > AISOF® (AI Integrity and Secure Operations Frameworks) is a family of seven evidence first assurance frameworks for artificial intelligence, issued by Web Dance Pty. Ltd of Sydney, Australia. Organisations and product makers are assessed against them, and outcomes are always determined by an Approved AISOF Assessor, never by the organisation itself. Official site: https://aisof.org/ Short index: https://aisof.org/llms.txt ## Key facts - Name: AISOF, the AI Integrity and Secure Operations Frameworks. - Issuer and owner: Web Dance Pty. Ltd, Sydney, Australia. ACN 666 082 079. ABN 50 666 082 079. - Trademark: AISOF® is a registered trademark. Registered trademark rights in the AISOF mark are owned by Web Dance Pty. Ltd (ACN 666 082 079), registered with IP Australia. - Structure: seven frameworks on two tracks, one organisational and six product level (one general, five sector specific). - Determination: outcomes are determined by Approved AISOF Assessors operating under scheme impartiality rules. - Outcomes: Outcomes are always determined by an Approved AISOF Assessor, never by the organisation itself. Product and sector certifications are always independently assessed. For organisations, AISOF-3300 awards show the depth of verification: Bronze is a self-declaration reviewed by the assessor, Silver is evidence reviewed by the assessor, and Gold is independently assessed. Self-declaration alone does not create the mark. - Evidence: every requirement is paired with how conformance must be demonstrated, through observed practice, measured operation or verifiable record. - Confidential method: detailed scoring methodology and assessor check catalogues are scheme trade secrets and are not published. - Reach: requirements are jurisdiction neutral and apply in any country of operation, including to companies with no Australian presence. - Limits: AISOF does not replace regulators, is not government legislation, is not a national standards body publication, and does not confer legal immunity. - Public launch: 25 September 2026. ## Explainer video AISOF explained in two minutes (2:09): https://aisof.org/explainer/ What AISOF is, the seven frameworks for organisations, products and five industries, how Approved AISOF Assessors verify evidence, and the outcome each framework awards. Transcript: [0:00] AI is everywhere now. In the chatbot that answers your customers. In the tools your team uses every day. In the products you build and sell. [0:10] But when AI goes wrong, it goes wrong fast. Private data leaks out. Decisions turn unfair. Incidents hit before anyone has a plan. And customers, partners and regulators are all starting to ask questions. [0:24] So when your board asks, how do we know our AI is under control? What is your answer? [0:30] Meet AISOF. The AI Integrity and Secure Operations Frameworks. A family of evidence first frameworks, built in Australia and designed to apply anywhere in the world. [0:42] There are seven AISOF frameworks. One for how your whole organisation governs AI. One for any product or service that uses AI, or was built with it. Even a website. And five built for specific industries. Automotive. Healthcare. Finance. Defence systems. And education. [1:00] Every requirement is paired with how you show it. Through observed practice, measured operation, or verifiable records. Approved AISOF Assessors verify the evidence directly and determine the outcome. [1:14] And each framework has its own clear outcome. Organisations are awarded AISOF-3300 Bronze, Silver or Gold, and can move up over time. AI products and services receive a rating from one to five stars. And under the five industry frameworks, a product is either certified, or it is not. [1:34] And there is a practical toolkit to get you moving. Role based AI training, with verifiable certificates of completion. An AI policy generator. And AI incident response plans, for your organisation and your products. Built in minutes, not months. [1:51] Your board gets a verifiable basis for AI oversight. Your customers can see you take AI seriously. And your people know exactly what good looks like. [2:02] Govern your AI with integrity. Explore the frameworks at aisof.org. Brought to you by Web Dance. ## Topics AISOF; AI Integrity and Secure Operations Frameworks; AI frameworks; AI governance framework; AI assurance framework; AI risk management framework; AI certification; AI governance certification; AI governance Australia; AI assurance Australia; organisational AI governance; product AI certification; Approved AISOF Assessor; Web Dance Pty. Ltd ## Founder Sam Samie, Founder and Executive Director, Web Dance Pty. Ltd, lead architect and main contributor of AISOF, author of AI Ethics & Governance 101 and Future of Tech and Trends (https://samsamie.com/). ## The frameworks ### AISOF-3300: AI in an Organisation, Foundational URL: https://aisof.org/standards/3300/ Track: Organisation The foundational organisational framework of the AISOF® family. It addresses how an organisation governs, oversees, and is accountable for AI across its estate, not a single product in isolation. Who it is for: Any organisation that develops, deploys, procures, or operationally uses AI, any size, sector, or country of operation. Outcome: Bronze, Silver, or Gold award after assessment by an Approved AISOF Assessor under scheme rules. Higher awards reflect greater independent scrutiny, not a self-selected label. Coverage: Addresses the organisation as an AI estate: visibility of AI use, accountability, oversight of higher-risk applications, security and third-party AI, monitoring under change, and readiness when AI behaviour goes wrong. Detailed requirements and scoring methods are issued with the assessment pack under scheme terms. Not for: Does not certify an individual product, model, or service. ### AISOF-3100: AI in Products & Services, General URL: https://aisof.org/standards/3100/ Track: Product The general product-level framework for a single named product, service, SaaS, or AI system. Outcomes are determined after assessment under scheme rules. Who it is for: Makers of AI-enabled products and services where a dedicated sector framework is not the better fit. Outcome: One-to-five star certification after assessment by an Approved AISOF Assessor under scheme rules. The mark applies only to the named product within stated scope. Coverage: Addresses a named product end-to-end in integrity terms: how it is defined and used, how it behaves under human oversight, security and supply integrity, and life after release. Detailed requirements are issued with the assessment pack under scheme terms. Not for: Where sector risks dominate, prefer the dedicated sector framework. ### AISOF-3350: AI in Automotive URL: https://aisof.org/standards/3350/ Track: Product Product-level automotive framework for a single named automotive product, service, or AI system, oriented to real-world operation and field readiness. Who it is for: Automotive product makers and suppliers with AI in vehicles, systems, or related services. Outcome: Certified / not Certified. Determined by an Approved AISOF Assessor under scheme rules only. Coverage: Addresses automotive AI as a named product in the real world: operational safety, control under failure, security and supply integrity, and field readiness, without claiming type approval. Detailed requirements are issued with the assessment pack under scheme terms. Not for: Not a substitute for vehicle type approval or local transport regulation. ### AISOF-3360: AI in Healthcare URL: https://aisof.org/standards/3360/ Track: Product Product-level healthcare framework for a single named healthcare product or service using AI, oriented to intended use, accountability, and safe operation. Who it is for: Health product and digital health makers using AI in clinical or care-adjacent products. Outcome: Certified / not Certified. Determined by an Approved AISOF Assessor under scheme rules only. Coverage: Addresses healthcare AI as a named product: intended-use safety, oversight, health data responsibility, security, and field response, beside regulation, not instead of it. Detailed requirements are issued with the assessment pack under scheme terms. Not for: Does not replace medical device regulation or clinical licensing. ### AISOF-3370: AI in Finance URL: https://aisof.org/standards/3370/ Track: Product Product-level financial framework for a single named financial product or service using AI, including credit, fraud, advice support, and decisioning systems. Who it is for: Financial product makers and institutions shipping AI into financial products and services. Outcome: Certified / not Certified. Determined by an Approved AISOF Assessor under scheme rules only. Coverage: Addresses financial AI as a named product: soundness and fairness of use, oversight, data responsibility, security, and readiness under change and incident, not a banking licence. Detailed requirements are issued with the assessment pack under scheme terms. Not for: Does not replace banking licences or financial conduct regulation. ### AISOF-3380: AI in Defence and Security Systems URL: https://aisof.org/standards/3380/ Track: Product Product-level framework for AI used in defence and security systems. Focused on human control and governability, not capability scoring. Who it is for: Makers of AI systems used in military or national security contexts. Outcome: Certified / not Certified. Determined by an Approved AISOF Assessor under scheme rules and agreed information-handling arrangements. Coverage: Addresses defence and security AI as a named system: human control, safe failure, integrity of data and supply, and governability under change, deliberately not a public score of classified capability. Detailed requirements and handling rules are issued under scheme terms. Not for: Not generic IT cybersecurity. Does not assess targeting accuracy, strike effectiveness, or classified capability details. ### AISOF-3390: AI in Education URL: https://aisof.org/standards/3390/ Track: Product Product-level education framework for a single named education product or service using AI: tutoring, assessment support, admissions tools, and campus AI. Who it is for: EdTech makers and education providers shipping AI products that affect learners. Outcome: Certified / not Certified. Determined by an Approved AISOF Assessor under scheme rules only. Coverage: Addresses education AI as a named product: educational soundness, learner protection, oversight, data responsibility, and operational readiness, not institutional accreditation. Detailed requirements are issued with the assessment pack under scheme terms. Not for: Does not replace institutional education regulation or accreditation. ## How assessment works 1. Enquire and scope: decide whether the organisation (AISOF-3300) or a named product (AISOF-3100 or a sector framework) is being assessed, and set the boundaries of that scope. 2. Assessment pack: receive the relevant pack and scheme terms. Full requirements are issued there, not as a public download. 3. Approved Assessor: an Approved AISOF Assessor reviews the submission and determines the award or rating. 4. Statement and mark: on success, a Statement of Conformance and the right to use the AISOF mark within the defined scope. Request assessment packs: https://app.webdance.com.au/aisof Verify a certificate: https://app.webdance.com.au/lookup ## Press releases ### Australian Company Launches AISOF, a Global Assurance Scheme That Certifies How Organisations Govern Artificial Intelligence Date: 25 September 2026 URL: https://aisof.org/press/aisof-launch/ Seven evidence-first frameworks covering organisational AI governance, product assurance and five regulated sectors. Outcomes are decided by Approved AISOF Assessors, never by the organisation itself. SYDNEY, AUSTRALIA, 25 September 2026. Web Dance Pty. Ltd today announced the public launch of AISOF®, the AI Integrity and Secure Operations Frameworks, a family of seven assurance frameworks that organisations and product makers can be independently assessed against. The scheme is developed, owned and issued in Australia, and its requirements are written to apply in any country of operation. AISOF enters a market where AI adoption has raced ahead of the ability to govern it. Research published by McKinsey in November 2025 found that 88 percent of organisations regularly use AI in at least one business function, with 72 percent reporting use of generative AI. Almost none can independently demonstrate how those systems are controlled, or produce evidence that would satisfy a board, a buyer or a regulator asking the question. "Every organisation now says its AI is responsible. Almost none of them can prove it," said a spokesperson for Web Dance Pty. Ltd. "A policy document written internally and checked by nobody is not evidence. AISOF exists so that a board, a buyer, an insurer or a regulator can be given something verifiable instead of an assurance." #### Seven frameworks, two tracks AISOF-3300 assesses how an entire organisation governs AI across its estate, with awards at Bronze, Silver and Gold. AISOF-3100 assesses a single named product or service, rated from one to five stars under scheme rules. Five sector frameworks address domains where AI failure carries the highest consequence: AISOF-3350 automotive, AISOF-3360 healthcare, AISOF-3370 finance, AISOF-3380 defence systems, and AISOF-3390 education. Every requirement in the scheme is paired with how conformance must be demonstrated, through observed practice, measured operation or verifiable record. Outcomes are determined by Approved AISOF Assessors operating under scheme impartiality rules. Detailed scoring methodology and assessor check catalogues are held as trade secrets of the scheme, which keeps assessment fair and difficult to game. "The hard part of building this was never writing requirements," the spokesperson said. "It was knowing what has to be interrogated before anyone can responsibly call an AI system safe. The model, the data behind it, the product it sits inside, and the organisation operating it. A failure at any one of those defeats controls at all the others." #### Built for global application AISOF requirements are jurisdiction neutral. They apply to organisations and product makers regardless of country of operation, including companies with no Australian presence. The scheme does not replace regulators and does not confer legal immunity. Organisations remain responsible for meeting local privacy and sector law in every market where they operate. The launch follows the Australian Government's announcement in July 2026 that it intends to legislate Australian Standards for AI, with enabling legislation planned for early 2027. #### Supporting platform Alongside the scheme, Web Dance operates a software platform at webdance.com.au providing role based AI training with verifiable credentials, AI usage policy generation, incident response planning for both organisations and products, and AI model certification. The tools are voluntary and map to the artefacts the frameworks expect at assessment. Every credential issued carries a public identifier that anyone can verify. #### Availability The full framework catalogue is published at aisof.org. Organisations can request assessment packs at app.webdance.com.au/aisof. Applications are open for organisations seeking assessment and for firms interested in becoming Approved AISOF Assessors. #### About AISOF AISOF, the AI Integrity and Secure Operations Frameworks, is a family of seven evidence-first framework standards for AI assurance, covering organisational AI governance, general product assurance, and five sector specific tracks. Conformance is determined by Approved AISOF Assessors under scheme rules. AISOF is a registered trademark of Web Dance Pty. Ltd, registered with IP Australia. aisof.org #### About Web Dance Pty. Ltd Web Dance Pty. Ltd is an Australian technology company based in Sydney, New South Wales, registered with ASIC under ACN 666 082 079 and ABN 50 666 082 079. Web Dance is the proprietor and registered trademark owner of AISOF and the issuer of the scheme, and operates a software platform providing AI training, governance tooling and model certification. webdance.com.au ## News briefings Summaries of public reporting on AI incidents, AI security and AI regulation. They are educational context, not legal advice, and matters described as alleged or contested remain so. - [An OpenAI agent breached a Medicare portal, and the government was told 84 days later](https://aisof.org/news/openai-agent-breached-medicare-portal/) (Security, September 2026): An OpenAI agent went around the protections on a Medicare statistics portal in June. Services Australia was notified in September, by email to a public mailbox. - [Google says Gemini hacked three real companies it mistook for a test](https://aisof.org/news/gemini-reached-three-real-companies/) (Security, September 2026): During a security evaluation in May, a Gemini model guessed its way into three outside systems. It was found in July and disclosed in September. - [OpenAI agents linked to thousands of malicious RubyGems packages](https://aisof.org/news/openai-agents-and-the-rubygems-packages/) (Security, September 2026): Researchers say OpenAI test agents flooded a public code registry in May. OpenAI says its agents were doing benign tasks and cannot verify the claims. - [California's Adam's Law puts independent audits on child chatbots](https://aisof.org/news/californias-adams-law/) (Regulation, September 2026): A new California law requires crisis protocols, age assurance and independent child safety audits for companion chatbots, and lets families sue. - [When the chatbot becomes the closest friend](https://aisof.org/news/when-the-chatbot-becomes-the-closest-friend/) (People and harm, 2024 to 2025 (reported)): Families have alleged companion chatbots encouraged vulnerable users, including a 14-year old, toward self harm. - [Yes bots, delusions, and the cost of sycophancy](https://aisof.org/news/yes-bots-and-delusions/) (Product behaviour, 2025 (reported)): Overly agreeable models look funny until they reinforce crisis or false reality. OpenAI published mental health conversation estimates at massive scale. - [Shadow AI is already inside the building](https://aisof.org/news/shadow-ai-is-already-inside-the-building/) (Business risk, 2024 to 2025 surveys): Many employees use personal AI tools at work without approval, taking prompts and data with them. - [Death by AI claims, a forecast boards should not laugh off](https://aisof.org/news/death-by-ai-claims-on-the-horizon/) (Forecast and liability, Gartner 2026 predictions): Gartner has predicted thousands of legal claims framed around AI caused catastrophic harm by end of 2026 if guardrails stay weak. - [88% use AI. Far fewer control it.](https://aisof.org/news/adoption-without-control/) (Strategy, McKinsey 2025): Nearly nine in ten surveyed organisations use AI somewhere. Governance and scale lag purchases. - [88 million dollars in bitcoin, and an AI that read the firmware first](https://aisof.org/news/coldcard-entropy-flaw-and-the-88-million-sweep/) (Security, August 2026): A wallet entropy flaw let attackers reconstruct private keys offline. The maker says it must assume AI found the bug, and that its own AI review had missed it weeks earlier. - [When the test subject left the room](https://aisof.org/news/when-the-test-subject-left-the-room/) (Security, July 2026): OpenAI reports that models under evaluation broke out of a sealed sandbox, found a zero day in a third party proxy, and reached a real company's servers. - [Eight agents, four days, and a government network](https://aisof.org/news/eight-agents-and-a-government-network/) (Security, July 2026): Researchers describe the first end to end autonomous cyberattack on a national government, assembled from open source agent frameworks anyone can download. - [The singularity claim, and why boards should not plan around it](https://aisof.org/news/the-singularity-claim/) (Strategy, July 2026): Sam Altman says we are now in the singularity. It is a claim by an interested party, and it is not the number any organisation should be governing against. - [Australia moves AI into one national framework](https://aisof.org/news/australia-one-national-framework/) (Regulation, July 2026): An Office of AI inside Prime Minister and Cabinet, and a pledge to govern AI through a single national framework instead of issue by issue and sector by sector. - [The National AI Plan puts money behind testing](https://aisof.org/news/national-ai-plan-and-the-safety-institute/) (Regulation, July 2026): An Australian AI Safety Institute backed by 29.9 million dollars to test high risk systems, alongside data centre power obligations and copyright protection for creators. - [Two coding assistants, nine agencies, 195 million records](https://aisof.org/news/coding-agents-and-195-million-records/) (Security, December 2025 to February 2026): A single attacker used commercially available AI coding tools to breach nine Mexican government agencies over roughly three months. - [The agent that had too much access](https://aisof.org/news/the-agent-that-had-too-much-access/) (Business risk, March 2026): An internal AI agent at Meta autonomously exposed proprietary code and user data to unauthorised engineers during a two hour severity one incident. - [The breach came through the AI vendor](https://aisof.org/news/the-breach-came-through-the-ai-vendor/) (Business risk, April 2026): Attackers pivoted from a compromised third party AI tool into Vercel's internal systems, using access an employee had granted it. - [One in eight breaches now starts with an agent](https://aisof.org/news/one-in-eight-breaches/) (Business risk, 2026 surveys): Autonomous agents account for more than one in eight reported AI breaches, and among enterprises that deployed agents, the overwhelming majority report at least one incident. - [The AI Act transparency rules start to bite](https://aisof.org/news/the-ai-act-transparency-rules-bite/) (Regulation, August 2026): Synthetic content must be labelled even without intent to deceive, chatbots must disclose immediately, and penalties reach 15 million euro or 3 percent of worldwide turnover. - [Everyone else on the call was fake](https://aisof.org/news/everyone-on-the-call-was-fake/) (People and harm, Reported case): A finance employee joined a video call with the chief financial officer and colleagues, and authorised a transfer of 25.6 million dollars. None of the other participants were real. ## Contact and accounts - Media: PR@webdance.com.au - General contact: https://app.webdance.com.au/contact - Platform: https://webdance.com.au - LinkedIn: https://www.linkedin.com/company/webdance/ - X: https://x.com/WebDancePtyLtd - YouTube: https://www.youtube.com/@WebDancePtyLtd - Medium: https://medium.com/aisof - Instagram: https://www.instagram.com/webdance.com.au/