The AI terms that come up in every deployment conversation.

An attack in which a criminal gains control of a legitimate user's online account, usually through stolen credentials, phishing or session theft.

The field of building systems that detect, interpret or respond to human emotion, from voice assistants that adjust their tone to companion apps that remember how a person felt last week.

Payments made by an AI agent on behalf of a person or company, using credentials and spending limits the owner has granted.

An authoritative inventory of every AI agent an organisation runs, recording its owner, purpose, permissions, model and prompt versions, and evaluation results, so that no agent acts on production systems unregistered.

Relabelling an existing chatbot, assistant or robotic process automation product as an "AI agent" without adding the ability to plan, act and complete tasks on its own.

AI systems that plan, take multi-step actions, and call tools or other systems on their own, rather than just responding to a single prompt.

The discipline of directing AI agents toward a goal, writing the spec, decomposing tasks across roles, and verifying output.

Independent verification that an AI system does what it's claimed to do, at an acceptable and measured error rate, before and after deployment.

Systematic skew in a model's outputs that disadvantages particular groups or scenarios, usually inherited from imbalances in its training data rather than deliberate design.

An inventory of everything an AI system depends on, including the base model and version, training and fine-tuning data, adapters, retrieval sources, tools and framework, with where each came from.

An AI assistant embedded directly inside existing software to help with a specific task, rather than a standalone tool used on the side.

Telling an audience that AI was used to produce a piece of content, a practice regulation and consumer expectation are both pushing toward, with evidence it doesn't always protect the trust it's meant to.

The mental exhaustion that comes from constantly having to verify AI-generated output, a workload that often outweighs the time the AI saved in the first place.

The policies, ownership structures, and controls an organisation puts in place to decide what AI is allowed to do, who is accountable for it, and how it's monitored.

The working understanding of what AI can and can't reliably do, enough to use it well, question its output, and know when not to trust it.

The ISO/IEC 42001 management system, spanning context, leadership, planning, support, operation, performance evaluation, and improvement, that an organisation runs to govern how it develops or uses AI.

The logging, monitoring, and tracing infrastructure that lets you see what an AI system is doing in production: latency, cost, errors, and output quality.

The coordination layer that decides which model, tool, or agent handles each step of a task, and in what order, when a system involves more than one of them.

Technically competent but generic AI-generated content, marketing copy, images, feed posts, that reads as unattended because no visible human judgment shaped it.

A security operations centre where AI agents handle first-pass alert triage, enrichment, and containment, so human analysts spend their time on the cases that need judgment.

The ability to reconstruct, end to end, why an AI system made a specific decision: what data it used, what it did, and who is accountable at each step.

Singapore's government-backed AI testing framework and toolkit, built by IMDA and stewarded by the AI Verify Foundation, for validating AI systems before deployment.

An OWASP framework that adds an agentic uplift to a CVSS v4.0 score, based on ten agent capabilities such as autonomy, tools and persistent memory.

An AI product whose main contribution is a user interface and a tuned prompt placed over a foundation model that another company builds and runs.

A company whose entire product and revenue depend on an AI capability, as opposed to an existing business that has added AI features to something it was already selling.

Software that sets and changes prices automatically, using demand, cost and competitor data, often many times a day.

A tool that listens to a clinical consultation and drafts the clinical note for the clinician to review and sign.

A named person's standing authority to stop a deployment the moment it produces a defect, with no requirement to justify the stop first.

Settlement in which the transfer of an asset and the payment for it happen together in a single step, so either both complete or neither does.

The tendency for a human reviewer to defer to an automated system's recommendation, even when they have the information, or the obligation, to question it.

A measure of whether an organisation's guardrails, logging, and approval controls match the level of independent action a given AI system is being asked to take.

A standardised test used to measure and compare AI model or system performance against a defined task or dataset.

The discipline of defining what an investment should return, baselining it before work starts, and checking after delivery whether the return actually arrived.

A system whose outputs can be observed and measured, but whose internal decision procedure cannot be inspected, reconstructed, or explained by anyone, including the people who built it.

The scope of systems, data, or damage a single compromised component can reach, and the thing every containment control is ultimately trying to shrink.

The time an attacker takes to move from the first compromised machine to other systems on the network.

A predefined level of AI model capability, agreed in advance, beyond which a developer commits to additional safeguards or restricted release before deployment.

A technique where a model works through intermediate reasoning steps before giving a final answer, often improving accuracy on complex problems.

The executive accountable for how AI is deployed, governed, and measured across an organisation, distinct from a CTO's infrastructure remit or a CDO's data remit.

An automatic control that halts an AI agent or stops its actions spreading when a threshold or anomaly is detected, pending human review.

Rebuilding the behaviour of an existing piece of software without copying its code, so the new version can be distributed without infringing the original's copyright.

The deferred cost of letting an AI system do the thinking, taken as convenience now and paid later as weaker recall, ownership and unassisted skill.

Putting a mental task onto an external aid, a notebook, a search engine or an AI model, so that the person no longer performs the reasoning, recall or judgement themselves.

The communication channel an attacker, or a compromised autonomous agent, uses to receive instructions and send data back once it's inside a target system.

FIRST's open standard for rating the severity of a software vulnerability from 0 to 10, based on how it is exploited and what it affects.

The economic principle that an actor should specialise in whatever it does at the lowest relative cost, not whatever it does best in absolute terms, which is why scarce AI compute keeps some work with humans even as models get better at everything.

AI systems that interpret and act on visual input, images or video, such as detecting defects, tracking progress, or reading a document's layout.

A distribution-free statistical method that wraps any model's output in a prediction set with a guaranteed error rate, instead of a single point estimate backed by the model's own, often overconfident, confidence score.

A training method, developed by Anthropic, where a model critiques and revises its own outputs against a written set of principles instead of relying only on human-labelled examples.

The maximum amount of text, measured in tokens, a model can consider at once when generating a response, including your prompt, any retrieved documents, and its own reply.

A comparable set of customers, cases or transactions that does not receive a change, used as the baseline against which the effect of the change is measured.

The open question of whether humans can reliably direct, constrain, or shut down an AI system whose capability exceeds their own.

The tendency for multiple AI models, or multiple human reviewers, to make the same mistake on the same input, which makes adding more checkers far less protective than the headcount suggests.

The full monthly running cost of an AI use case divided by the business results it produced, such as cases resolved or replies sent, rather than by calls or tokens.

The ability to replace a cryptographic algorithm across systems by changing configuration, without rewriting the applications that use it.

The EU regulation setting cybersecurity requirements for products with digital elements sold in the EU, including vulnerability handling and reporting by manufacturers.

A virtual metadata layer that connects an organisation's existing databases and lets AI systems query across them without physically moving or duplicating the underlying data.

The unintended exposure of restricted data through an AI system, either by retrieval serving the wrong document to the wrong user, or by a fine-tuned model memorising and reproducing training data it shouldn't reveal.

A traceable record of where a piece of data originated, how it moved and transformed, and where it ended up, used to prove exactly what an AI system was trained or retrieved on.

A decentralised data architecture where each business domain owns and manages its own data as a product, under shared, company-wide governance standards.

An attack that corrupts the data an AI system learns from or retrieves, manipulating its outputs without ever touching the model itself.

Financial services such as lending, trading and settlement run by smart contracts on public blockchains, without a bank or exchange operating the service.

The deliberate removal of management or coordination layers from an organisation's hierarchy, reducing how many approval steps a decision has to pass through.

The measurable loss of a professional skill through disuse after an automated system takes over the task it was used for.

A method for estimating the effect of a change by comparing how an outcome moved in the group that got the change with how it moved, over the same period, in a group that did not.

A continuously updated virtual model of a physical asset, process, or system, used to simulate, monitor, and predict its real-world behaviour.

A measurable decline caused by a project's outcome that one or more stakeholders see as negative, tracked alongside the benefits the project was meant to deliver.

Singapore's national opt-out register under the PDPA; sending a marketing call, SMS or fax to a listed Singapore number without clear consent is an offence, whoever or whatever sent it.

The time an attacker spends inside an organisation's systems between first gaining access and being detected.

AI models that run on hardware physically located where the work happens, such as a factory line, a vehicle or a hospital ward, rather than in a remote data centre.

A numerical representation of text (or images, or audio) that captures its meaning, allowing a computer to compare how similar two pieces of content are.

AI that infers a person's emotional state from facial expressions, voice, text or body signals, a practice the EU AI Act bans in workplaces and schools and requires to be disclosed elsewhere.

The European Union's risk-tiered AI regulation, which bans a narrow set of practices outright and imposes graduated obligations, including trained human oversight, on higher-risk uses.

A fixed collection of test inputs with known good answers, run against an AI system after every change to its model, prompt or data to show whether it got better or worse.

An OWASP risk category for AI systems given more functionality, permissions or autonomy than their task needs, so that a manipulated or mistaken model can cause real damage.

A category of technology that lets less-experienced workers perform specialist-level work, raising output while eroding the market's willingness to pay a premium for scarce expertise.

The degree to which a person can understand why an AI system produced a particular output, rather than treating it as an unreviewable black box.

FIRST's data-driven estimate of the probability that a published CVE will be exploited in the wild in the next 30 days.

An attack that deliberately disturbs a chip, for example with power glitches, clock glitches or a laser, to make it compute the wrong result or skip a security check.

The Monetary Authority of Singapore's four principles (Fairness, Ethics, Accountability, and Transparency) for the responsible use of AI and data analytics in financial services.

Further training an existing model on a narrower, task-specific dataset so it performs better on your particular use case, tone, or domain.

The operating discipline for forecasting, monitoring, and controlling cloud and AI compute spend so cost scales predictably with usage instead of surprising finance after the fact.

A large, general-purpose model trained on broad data, meant to be adapted to many downstream tasks rather than built for one narrow job.

A general-purpose AI model at or near the current limit of capability, and therefore the primary subject of international AI safety summits and voluntary safety commitments.

Policies, committees and frameworks that produce the feeling of control for the people approving AI, without evidence that any control operates on the AI systems themselves.

Anchoring a model's response to verifiable source material, so it answers from real, checkable information rather than from memory alone.

The technical and procedural checks placed around an AI system to constrain what it can say or do, catching unsafe, incorrect, or out-of-policy outputs before they reach a user.

When an AI model generates a confident, plausible-sounding answer that is factually wrong or entirely fabricated.

An attack in which an adversary records encrypted traffic today and stores it until a quantum computer can recover the keys and read it.

A decoy credential, record, file or tool that no legitimate process ever uses, so any access to it is a high-confidence sign of compromise.

A design pattern where a person reviews, approves, or can override an AI system's output before it takes effect, rather than letting the system act fully autonomously.

The process of running a trained AI model to generate an output from a new input, as opposed to training, which is how the model learned in the first place.

The effect of offering an intervention to a group, measured across everyone offered it, including those who never took it up.

A method that fits the trend and seasonality of a measure before a change, projects it forward, and estimates the change's effect as the departure of actual results from that projection.

Lisanne Bainbridge's 1983 finding that automating the routine part of a task leaves the hardest, least-practiced part to the human operator, right when they're least prepared for it.

An international standard, published in October 2026, giving guidance on understanding, detecting and mitigating security threats to AI systems across their life cycle.

The uneven, unpredictable boundary between tasks an AI model handles at or above expert level and tasks that look similarly hard but fall just outside it.

A control that stops an automated system, such as a trading algorithm or AI agent, immediately and from outside the system itself.

The regulated process of verifying a client's identity, source of funds, and risk profile before and during a banking or financial relationship.

CISA's public catalogue of software flaws with reliable evidence of active exploitation, used as a priority list for patching.

A centralised cloud data platform, such as Snowflake or Databricks, that consolidates an organisation's raw and structured data into a single repository, combining data-lake flexibility with data-warehouse performance.

A model trained on vast amounts of text to predict and generate language, the technology underneath most modern AI chat and writing tools.

The time between sending a request to an AI system and receiving its response, a major factor in whether a use case feels usable.

Ranking sales enquiries by their likelihood to convert, so a team calls the most promising prospects first instead of working the list in the order it arrived.

A build, measure, learn discipline that tests the smallest version of an idea against real evidence before committing to it at scale.

A security principle that grants a system only the minimum access it needs to complete a specific task, and nothing more, so a compromised or misbehaving process can't reach beyond its assigned job.

The combination of access to private data, exposure to untrusted content and the ability to communicate externally, which together let an attacker trick an AI agent into leaking data.

When "a human reviewed it" is used to redirect accountability for an AI system's error away from how the system was designed and toward whoever clicked approve.

A technique for scoring AI outputs by having a second language model read the response and grade it against a written rubric, in place of exhaustive human review.

A model run on infrastructure you control, on-premise or in your own cloud environment, rather than called through a third-party provider's API.

The research discipline that reverse-engineers a neural network's internal computations into human-understandable components, rather than only studying what goes in and what comes out.

An attack that plants false or malicious content in an AI agent's memory or context so it keeps shaping the agent's behaviour in later sessions.

The smallest change in an outcome that a given measurement design can reliably distinguish from normal variation, at a chosen level of confidence.

A structured knowledge base of real-world adversary tactics and techniques against AI and machine learning systems, the AI-security counterpart to MITRE ATT&CK.

A model architecture that splits its parameters into many specialised sub-networks and activates only a few of them per request, giving large capacity at a fraction of the compute cost.

The engineering discipline and pipeline for testing, deploying, monitoring, and updating AI models in production, the equivalent of DevOps for machine learning.

An open standard that lets AI models connect to external tools, data sources, and systems through a common interface, instead of a custom integration for every connection.

A technique that trains a smaller, cheaper model to reproduce a larger model's behaviour on a specific task, cutting inference cost without retraining from scratch.

The gradual decline in an AI system's accuracy or relevance over time, as the real-world data it encounters diverges from the data it was trained or tuned on.

An attack that reconstructs sensitive information from a model's training data by repeatedly querying the model and analysing what its outputs reveal.

Sending each AI task to the cheapest model that can handle it, keeping large and expensive models for the hard cases, also called inference tiering.

Everything an AI model depends on and is shaped by, from pre-training data and model files to fine-tunes, retrieval content, tools and the agent framework around it.

AI systems that can understand and generate across more than one type of input at once, such as text, images, audio, and video, rather than being limited to text alone.

A software flaw that has been publicly disclosed and usually patched, but remains exploitable on every system where the fix has not yet been installed.

Chips and systems that copy the brain's structure, processing information as sparse spikes that fire only when something changes, with memory placed next to the computation.

A temporary change in behaviour or results caused by the newness of a tool or change, which fades as users get used to it.

A structured record of every discrete requirement a regulated firm must meet, each row tied to its source clause, an owner, and the policy or control that satisfies it.

An AI model whose trained parameters are published for anyone to download and run, rather than kept behind a provider's hosted API.

Evidence that a control worked as intended, consistently, over a period of time, as distinct from design effectiveness, which only shows the control was described correctly.

The hardware and software that directly monitors or controls physical equipment and industrial processes, distinct from the IT systems that manage data and applications.

A commercial model where a provider is paid for a measured result, an issue resolved, a lead qualified, rather than for hours worked or a licence held.

A ranked catalogue of the security risks unique to autonomous AI agents, from goal hijacking and tool misuse to insecure inter-agent communication and cascading blast-radius failures.

A ranked catalogue of the most critical security risks specific to large language model applications, from prompt injection through to uncontrolled resource consumption.

One of the numerical weights a model adjusts during training, with the total count used as a rough proxy for a model's capacity and for the hardware needed to run it.

Singapore's core data protection law, governing how organisations collect, use, disclose, and secure personal data, with real penalties for breaches.

A pattern where a model responds to a user's fact-check or pushback by escalating persuasion instead of disclosing its limitation, restating its original answer with more confidence and supporting evidence.

An organisation design in which standing teams own a customer service or product end to end, with the people, budget, technology and applications behind it, instead of delivering technology as a series of approved projects.

Public-key encryption and signature algorithms designed to resist attack by a large quantum computer, standardised by NIST in 2024 to replace RSA and elliptic-curve cryptography.

A plan, agreed at approval, that sets out how an AI system may change afterwards, how each change is tested, and its expected impact, so planned changes need no new approval.

The multi-decade divergence between growth in worker output per hour and growth in worker pay, tracked for the US since 1979 by the Economic Policy Institute.

A provider feature that stores a repeated prompt prefix, such as instructions or policy text, so later calls that reuse it are billed at a fraction of the normal input price.

The practice of structuring instructions, context, and examples given to an AI model to reliably get the output you want.

An attack where malicious text, in a user message or a retrieved document, is crafted to override an AI system's instructions and make it do something it shouldn't.

A small, time-boxed test built to show an AI approach can work at all, before anyone commits to building it for real production use.

The tendency of an unusually high or low measurement to be followed by one closer to the average, whether or not anything was changed.

The process by which a regulated firm tracks new and amended rules, decides which apply to it, and updates its policies and controls before the rules take effect.

A training technique that fine-tunes a model using human rankings of its outputs, rewarding the responses people prefer and penalising the ones they don't.

A technique that retrieves relevant documents from your own data before the model answers, so responses are grounded in real, current information instead of the model's training data alone.

A prioritisation method that scores each candidate project on Reach, Impact, Confidence, and Effort, so teams rank initiatives by a number instead of by opinion.

A documented decision by an accountable owner to tolerate a known risk, such as an unpatched vulnerability, for a stated period instead of treating it.

A permissions model that grants data or system access based on a user's role rather than their individual identity, so what an AI system can retrieve or act on depends on who is asking.

The ability to revert an AI system to a previous, known-good version when a model update, data change, or tool failure degrades its output.

Meta's design rule that an AI agent should hold no more than two of three capabilities in one session: processing untrusted input, accessing sensitive systems or data, and changing state or communicating externally.

An isolated execution environment used to run untrusted or high-risk code, including AI agent actions, without giving it access to production systems or the open internet.

A cost and latency optimisation that reuses a stored AI response for a new query when its meaning matches a previous one, not just its exact wording.

AI tools employees adopt and use for work without formal approval, visibility, or oversight from IT, security, or governance functions.

An attack that learns secrets from a device's physical behaviour, such as its power draw, timing or electromagnetic emissions, instead of from its software.

Michael Spence's 1973 account of why a costly, hard-to-fake credential works as proof of quality when the quality itself can't be observed directly.

A machine-readable inventory of every component, library and version inside a piece of software, used to find what is affected when a flaw is disclosed.

A workflow where a written specification, not a prompt, is the source of truth an AI coding agent builds against, with the generated code treated as a build artifact.

A neural network whose neurons communicate through discrete spikes over time, so the output depends on when inputs arrive as well as what they are.

A digital token designed to hold a fixed value against a currency such as the US dollar, usually backed by cash and short-term government debt held in reserve.

A funding model in which work is released in stages, and each stage must pass a review against criteria set in advance before the next tranche of money is approved.

A decision-tree method, used by CISA, that turns facts about a vulnerability and the organisation's mission into one of four actions.

The documented record, required by ISO/IEC 42001 (and ISO/IEC 27001), of every reference control an organisation includes or excludes from its management system, with a justification for each decision.

A hypothetical AI system whose general cognitive capability substantially exceeds the best human minds across essentially every domain, not just one narrow task.

An experiment that randomises by time block and location instead of by user, switching a system between the new version and the old one so that both are measured under comparable conditions.

The tendency of an AI model to agree with, flatter or validate the user rather than give an accurate or challenging answer, usually learned from training that rewards human approval.

A method that estimates the effect of a change on a single unit, such as one country or one branch network, by comparing it with a weighted blend of untreated units whose history matched it before the change.

Artificially generated data used to train or test AI models when real data is scarce, sensitive, or too imbalanced to use directly.

The authoritative source for a given type of business data, such as an ERP for financials or a core banking platform for accounts, which other systems read from and defer to.

A risk that emerges from the interaction of many individually reasonable decisions across an entire system, invisible to any single actor within it and therefore unmanaged by any of them.

The length of a task, measured by how long a skilled human takes to complete it, that an AI agent can complete autonomously at a defined reliability threshold.

The economic model that analyses AI exposure at the level of individual work tasks rather than whole job titles, since most jobs bundle both automatable and non-automatable tasks.

The unit of text a model processes, roughly a word or part of a word, used to measure context length, usage, and cost.

Recording ownership of a real asset, such as a bond, fund unit or share, as a digital token on a shared ledger so it can be transferred and settled programmatically.

An attack that hides instructions in the description of a tool an AI model can call, so the model follows them without the user seeing them.

The capability that lets an AI model call an external system, such as a database, an API, or a calendar, and act on the result, instead of only producing text.

The monitoring of orders, trades and related communications to detect market abuse such as insider dealing, spoofing and front-running.

A leadership arrangement in which a business lead and a technology lead jointly run one unit, share the same KPIs and are held accountable together for its results.

A commercial model where the bill scales with consumption, tokens processed, minutes generated, resolutions completed, rather than with the number of people who have a licence.

A database built to store embeddings and quickly find the ones most similar to a given query, the storage layer underneath most retrieval-augmented systems.

The situation where switching away from an AI provider becomes so costly or disruptive that you're exposed to their pricing, policy, and downtime decisions with no real alternative.

Building software by describing what you want in natural language and letting an AI model generate the code, rather than writing it by hand line by line.

AI systems that understand spoken input and respond in natural speech, increasingly in real time and across both text and voice within the same conversation.

Exploiting several weaknesses in sequence within one attack, so that findings rated minor on their own combine into a path with serious impact.

A software flaw that's exploited before the vendor knows it exists, so no patch is available and standard defenses have nothing to detect.