Last updated · By the Clerk2Officer editorial team · Report a correction
Changing ethical dynamics
Explain new technology or market change.
P2D05: Banking Ethics: Changing Dynamics
Learning purpose
Technology changes the scale, speed and visibility of banking decisions. A decision once made for one customer can now be embedded in software and repeated thousands of times before anyone notices a problem. Ethical banking therefore requires attention to data, automated judgement, access, intellectual property and security throughout a service's life. This chapter explains how familiar duties of honesty, fairness and care apply when the decision-maker works through digital systems and external technology partners.
- Automation can scale both useful service and harmful decisions.
- Ethical responsibility extends through design, deployment, monitoring and retirement of technology.
Reference date: 16 October 2026. The DPDP Act, its commencement notification G.S.R. 843(E), and the DPDP Rules notification G.S.R. 846(E) were read separately. The November 2025 notifications provide immediate, one-year and eighteen-month phases; the latter phases remain future at this reference date. Current sectoral obligations must not be replaced by a premature assertion that every DPDP provision is already operative. Original ethical guidance below is distinguished from specific enacted provisions and their commencement.
1. Ethics and technology: who remains responsible?
A technological system reflects choices about objectives, inputs, rules, exceptions and access. Describing an outcome as “the computer's decision” does not explain who selected the data, approved the model or decided how customers could challenge an error. A bank should identify accountable owners and provide staff with enough understanding to use the system responsibly. Technical complexity can justify specialist support; it cannot justify treating harmful outcomes as nobody's responsibility.
- System outcomes reflect human and organisational design choices.
- Complexity calls for competent oversight and accountable ownership.
Digital convenience can create pressure to remove every point of friction. Some friction is unnecessary, such as repeated requests for the same properly available information; other friction protects the customer, such as a clear confirmation before a consequential instruction. Ethical design distinguishes the two. A process that increases completion by confusing customers about charges or making refusal difficult may improve a conversion metric while weakening informed choice.
- Remove unnecessary burden while preserving meaningful confirmation and understanding.
- A higher completion rate does not prove that customers made informed choices.
Accessibility should be considered from the beginning. A mobile-only journey may exclude customers with limited connectivity, disabilities, unfamiliarity with the language or difficulty using a device. Appropriate assistance and alternative routes can make a service more inclusive without asking customers to surrender credentials. Staff should not treat a customer's need for support as proof of incapacity or as a reason to pressure them into a service they do not understand.
- Accessible service requires attention to different abilities and circumstances.
- Assistance should preserve the customer's control and authentication security.
2. Data security and privacy are related but distinct
Data security protects information and systems against unauthorised access, alteration, loss or disruption. Privacy concerns whether personal information is collected, used and shared appropriately in relation to the person concerned and the applicable framework. A database can be strongly encrypted yet used for an unjustified purpose. Conversely, a legitimate purpose does not excuse inadequate security. A sound banking service needs both an appropriate basis for processing and effective protection of the information.
- Security concerns protection; privacy also concerns the appropriateness of processing.
- Encryption does not by itself justify collecting or using personal information.
Personal data can reveal more than a single visible field suggests. A series of transactions may indicate health expenses, relationships, location patterns or financial distress. Combining apparently ordinary records can create a much more revealing profile. Staff should therefore evaluate the purpose and consequences of combining datasets, rather than assume that information is harmless because each separate item seems routine or because the bank already holds it for another service.
- Combined records can reveal sensitive patterns beyond individual fields.
- Existing possession does not automatically justify every new use or combination.
Data minimisation is a useful ethical design principle: collect and expose what is needed for the legitimate task rather than everything technically available. If a demonstration needs only transaction patterns, a realistic synthetic dataset may be sufficient. If an operational task needs a subset of fields, staff should not send a full customer record merely because exporting it is easier. Reducing unnecessary data also reduces what can be exposed in an error or compromise.
- Match information collection and access to the task's actual need.
- Synthetic or limited data can reduce unnecessary exposure in demonstrations and testing.
Illustration: A support team needs four fields to investigate a service issue, but a routine export includes 20 fields for 500 customers. The export contains 10,000 field values, while the stated need covers 2,000. The remaining 8,000 are unnecessary under the example's assumptions. This arithmetic is not a complete privacy-risk measure, but it illustrates how a convenient export can multiply exposure without adding value to the task.
- The stated export contains 20 × 500 = 10,000 values; the need is 4 × 500 = 2,000.
- Unnecessary values total 8,000 under the assumptions, although risk also depends on their nature.
3. Purpose, consent and customer understanding
A customer should be able to understand the relevant purpose of a request and the practical choice being made. A broad phrase such as “for better service” can conceal unrelated uses unless it is explained meaningfully. Where consent is the applicable basis, design should support a real decision rather than rely on confusing language or bundled permissions. The legal basis must be identified accurately: it is also wrong to claim that every lawful processing activity depends on consent in every circumstance.
- Explain the actual purpose and identify the applicable basis for processing.
- Consent should not be treated either as a meaningless click or as the only possible legal basis universally.
Withdrawal or refusal should be explained without exaggeration. A customer declining an optional marketing use should not be told, without a valid basis, that all banking services must end. At the same time, withdrawal of a permission does not necessarily erase records that must be retained under an applicable obligation or cancel an existing contract. Staff need a clear distinction between optional future use, necessary service processing and lawful retention rather than an absolute promise of immediate universal deletion.
- Distinguish optional use, necessary processing and legally required retention.
- Do not promise that withdrawal automatically cancels contracts or erases every record.
Data accuracy is an ethical issue because incorrect information can influence lending, fraud controls and customer access. A wrong address, duplicated liability or mistaken identity can repeatedly produce adverse outcomes if the system treats the record as unquestionable. The bank should provide an appropriate correction and review route, trace how the error affected decisions and communicate honestly about the correction. Correcting one screen may not resolve copies or downstream uses automatically.
- Inaccurate data can cause repeated harm through automated decisions.
- Correction should consider downstream effects and relevant copies, not only one display.
4. DPDP: understand the framework and its phased commencement
The Digital Personal Data Protection Act, 2023 defines the Data Principal as the individual to whom personal data relates, subject to its representative provisions, and the Data Fiduciary as the person determining the purpose and means of processing. A Data Processor acts on behalf of a fiduciary. These roles are functional: a vendor's contractual description should be examined against what it actually does. They help explain responsibility, but a definition's commencement does not mean that every substantive obligation commenced at the same time.
- Principal, fiduciary and processor describe different roles in personal-data processing.
- Role definitions and substantive obligations can have different commencement dates.
G.S.R. 843(E), dated 13 November 2025, appoints publication as commencement for specified provisions, including section 2 and sections 18 to 26. It places section 6(9) and section 27(1)(d) in the one-year group, and major processing, consent, fiduciary-duty and individual-right provisions in the eighteen-month group. As at 16 October 2026, neither deferred period has elapsed. It would therefore be inaccurate to teach all of the Act's core processing obligations as already operative solely because the Act was passed in 2023.
- The Act notification expressly uses immediate, one-year and eighteen-month groups.
- The one-year and eighteen-month groups are future at the chapter's October 2026 reference date.
The Rules have their own commencement allocation. Rule 1 brings rules 1, 2 and 17 to 21 into force on publication; rule 4 is deferred for one year; rules 3, 5 to 16, 22 and 23 are deferred for eighteen months. The distinction matters for breach reporting: the detailed rule 7 regime cannot simply be substituted for existing RBI or CERT-In reporting duties before its commencement. Module C explains the separately applicable cyber-reporting triggers and clocks.
- Rules commencement must be checked separately from Act commencement.
- A future DPDP breach-reporting rule does not displace current sectoral reporting duties.
The future framework is still relevant to preparation. Banks can map data, review notices, identify processors, plan customer-rights processes and test security arrangements before a provision becomes operative. Preparation should be described honestly as readiness where that is its status. Ethical duties and existing applicable laws do not disappear during a transition, while future legal provisions should not be misrepresented to customers as an already effective source of a particular right or deadline.
- Readiness work can begin before a deferred provision takes effect.
- Transition requires accurate distinction between current obligations, future requirements and ethical good practice.
5. Automated decisions, fairness and explainability
A model can reproduce disadvantages present in its data or create new ones through a poorly chosen objective. For example, a variable that appears neutral may act as a proxy for an irrelevant personal characteristic. A bank should test performance across relevant groups, investigate unjustified differences and consider whether the data and objective are suitable for the actual decision. Merely removing an obvious sensitive field does not prove that the remaining model is fair.
- Bias can enter through data, objectives and proxy variables.
- Removing one field is not a complete fairness assessment.
Illustration: In a hypothetical review sample, 200 comparable applications from group A include 20 erroneous rejections, while 100 from group B include 20. Error rates are 10% and 20%, respectively. Equal error counts therefore hide different rates. The difference calls for investigation; it does not by itself prove unlawful discrimination or reveal its cause. Review should examine comparability, sample quality, reasons for errors and the proposed remedy.
- Equal counts can conceal different rates: 20/200 = 10%, while 20/100 = 20%.
- A disparity is an investigation signal, not a complete causal or legal finding.
Explainability means being able to provide a meaningful account of how a decision was reached at the level appropriate to the audience. A customer may need understandable reasons and a correction route, while a validator needs deeper technical evidence. These needs do not require publication of every security-sensitive detail or proprietary model component. A claim of trade secrecy should likewise not become a blanket excuse for refusing all accountability or review of potentially harmful decisions.
- Explanations should be meaningful for the customer, operator or reviewer concerned.
- Protecting legitimate secrets and providing accountable review are compatible objectives.
Human oversight is useful only when the person has competence, time, relevant information and authority to intervene. A reviewer who routinely clicks approval because the queue is too large adds little protection. The bank should define when escalation is required, what evidence the reviewer examines and how overrides are recorded and analysed. Human involvement is a control design problem, not a ceremonial signature attached to an otherwise unexplained automated process.
- Effective oversight requires capability and authority, not merely a human name on the workflow.
- Overrides and escalation decisions need evidence and review.
6. Intellectual property rights in banking
Intellectual property rights protect particular kinds of creations and interests under their respective legal frameworks. Copyright concerns protected expression, such as original software code or training material; patents concern qualifying inventions; trade marks identify the source of goods or services; confidential know-how may be protected through appropriate legal and contractual duties. These are not interchangeable rights. A bank should identify the relevant asset and permitted use rather than assume that paying for one copy transfers every right.
- Different intellectual property regimes protect different interests.
- Purchasing access or a copy does not automatically transfer ownership of all rights.
Official Indian copyright material includes computer programmes within literary works. A software licence may permit use under specified conditions without allowing unlimited copying, redistribution or modification. Similarly, a paid report or training manual should not be reproduced wholesale for another audience merely because an employee can download it. Attribution is good practice, but crediting the source does not by itself supply permission for every use or replace an applicable licence or legal exception.
- Computer programmes are included within literary works under Indian copyright law.
- Attribution and permission are different questions; licence terms still matter.
Open-source software is not the same as software without conditions. Different licences can impose requirements concerning notices, attribution, distribution, source availability or combinations with other software. The bank should review the actual licence and its intended deployment, especially when modifying or distributing components. A developer's statement that code is “free on the internet” does not establish that the bank may use it in any way or that it is secure and maintained.
- Open-source permissions are governed by the applicable licence.
- Availability without payment does not prove unrestricted use, security or support.
Case: A team obtains a tool licensed for 50 named users but gives access to 65 under the assumption that the extra staff use it only occasionally. On the stated licence terms, 15 additional users require resolution through procurement or an authorised licence arrangement. The team should not share credentials to disguise the position. The example concerns a hypothetical named-user contract; actual licence metrics may instead depend on devices, concurrency or other terms.
- Under the stated named-user terms, the excess is 65 − 50 = 15 users.
- Licence compliance depends on the actual metric and cannot be cured by concealing use.
7. Patents and proprietary rights
A patent concerns an invention meeting the applicable legal requirements, including novelty, inventive step and industrial applicability, subject to exclusions. It is not awarded simply because a product is commercially useful or has a modern name. A patent application is also not the same as a granted right, and a grant should not be described as permission to ignore every other legal requirement. A banking technology proposal needs competent intellectual-property review where patent questions are material.
- Commercial usefulness and branding do not by themselves establish patentability.
- Application, grant, ownership and lawful use are separate questions.
Section 3(k) of India's Patents Act excludes mathematical or business methods, computer programmes per se and algorithms. The final 2025 computer-related invention examination guidelines discuss how claimed inventions should be assessed, including the significance of technical substance and effects. The correct lesson is not that every software-related invention is patentable or that every invention using software is automatically excluded. Claims must be examined under the actual statutory requirements and applicable guidance.
- Section 3(k) requires careful analysis of the claimed subject matter.
- Avoid both universal approval and universal rejection of software-related inventions.
Ownership should be resolved explicitly in development contracts. A bank commissioning software may need rights to use, modify, maintain or transfer the work if the supplier fails, but those rights should not be assumed from the payment alone. Employee-created work, pre-existing vendor components and third-party libraries can raise different issues. Clear agreements, documented provenance and specialist review help prevent a critical service from depending on rights the bank does not actually possess.
- Development contracts should identify ownership and the rights needed for continuity.
- Pre-existing and third-party components can have different rights from newly commissioned work.
Proprietary protection also requires responsible conduct by the bank. Employees should not copy a competitor's confidential material, misrepresent the origin of an invention or claim a supplier's work as their own. Conversely, confidentiality clauses should be used for legitimate protection, not to conceal misconduct from a competent authority. Intellectual property and information ethics both depend on accurate attribution, authorised access and respect for the rights and obligations attached to the material.
- Respect others' protected work and represent provenance accurately.
- Legitimate confidentiality should not become a pretext for obstructing authorised scrutiny.
8. Ethics of information security
Security staff may have powerful access to systems and records. Their authority should be limited to approved purposes and scope, with suitable monitoring and separation of duties. Curiosity about a colleague's salary or a public figure's account is not a security investigation. Privileged users should be especially careful because their actions can expose large volumes of information or alter evidence that others rely on to detect misuse.
- Privileged access increases responsibility; it does not create unrestricted permission.
- Security work must remain within an authorised purpose and scope.
Vulnerability testing requires clear authorisation, defined targets, timing, safeguards and reporting arrangements. A staff member who discovers a weakness should report it through the approved route and avoid expanding access merely to prove how much can be taken. Responsible handling protects customers while enabling correction. Publicly revealing sensitive exploit details before appropriate handling can increase harm, while indefinite concealment from the competent owner can leave the weakness unresolved.
- Testing should have explicit scope, safeguards and a reporting route.
- A discovered weakness should be handled to support correction while limiting avoidable exposure.
Security monitoring should also be proportionate. Collecting every possible employee communication may exceed what is justified for the stated purpose, while insufficient monitoring can leave serious misuse undetected. The institution should establish legitimate objectives, appropriate access, retention and oversight, and communicate the applicable policy accurately. Ethical evaluation asks whether the intrusion is necessary and proportionate, rather than assuming either that all monitoring is improper or that security justifies unlimited surveillance.
- Monitoring needs a legitimate purpose and proportionate scope.
- Security objectives do not automatically justify unlimited collection or access.
9. Incident honesty and technology procurement
When a data incident occurs, the bank should preserve evidence, contain harm, make required reports and communicate through authorised channels. It should distinguish confirmed facts from estimates and correct earlier information when needed. Concealing an incident to protect reputation can delay protective action and deepen harm. At the same time, speculation and unnecessary publication of sensitive details can also be harmful; responsible communication is accurate, timely and appropriate to the audience and obligation.
- Incident communication should distinguish verified facts from uncertainty.
- Honest reporting and controlled disclosure should work together to reduce harm.
Combined case: A vendor offers an AI service that requires full customer statements, claims ownership of all resulting profiles and provides no clear way to investigate adverse outcomes. The bank should examine necessity, authority, data handling, contractual rights, model performance, security and customer redress before deployment. A successful demonstration does not settle those questions. An ethical procurement decision connects the proposed benefit with evidence that the bank can operate and supervise the service responsibly over time.
- Evaluate data needs, rights, explainability, security and redress alongside model performance.
- A demonstration is evidence of selected functionality, not proof of complete suitability or governance.
Revision points
- Technology changes how responsibility is exercised, not whether responsibility exists.
- Security and privacy are related but distinct; purpose and protection both matter.
- Minimise unnecessary data, correct errors and explain choices accurately.
- DPDP Act commencement and Rules commencement have separate phased allocations.
- Future DPDP requirements do not displace current applicable RBI or CERT-In duties.
- Automated fairness requires testing, meaningful review and effective human oversight.
- Copyright, patents, trade marks, licences and confidential know-how are distinct concepts.
- Privileged access, testing, monitoring and incident communication require authorised and proportionate conduct.
Chapter practice
This practice is separate from timed mock tests. Answers and explanations appear after you choose an option.
Your reading progress
Sign in to save your progress across devices.