On July 14, 2026, CNBC reported on a lawsuit against Meta regarding its recent layoffs in which former employees claimed Meta used metrics “like token consumption, which has become a proxy for general AI usage, in a way that targeted certain employees. …
“The lawsuit comes nearly a month after a federal judge in California ruled against tech firm Workday in a separate employee-related lawsuit involving the use of AI for hiring decisions. In that case, the judge ruled that Workday must face claims about the company’s use of AI-powered job screening services that allegedly violated state and federal laws pertaining to employee discrimination.”
And these two examples are just the tip of a growing iceberg of legal cases surfacing regarding AI usage in all areas of business within the public and private sectors.
To be fair, a Meta spokesperson told CNBC in an email that the “claims lack merit and are not based on facts”:
“‘Workforce management and organizational decisions were and are made by people, not AI,’ the Meta spokesperson said in the statement.”
HUMAN DECISIONS IN AN AI WORLD
Last week, in Part 1 of this three-part blog series, we looked at the importance of AI ethics overall as well as why prompt engineering needs a moral compass.
Digging deeper as we head into solutions in Part 2, what is becoming clearer across America in mid-2026 is that traditional corporate and government ethics policies are being applied whether AI is used in decision-making or not. Like a human behind the wheel of an autonomous vehicle, the car might be doing all the work, but human oversight is required and responsible for the outcomes.
(Note that for this piece, I am leaving aside the many thorny questions regarding truly autonomous vehicles or AI systems running solely on their own. “Human-in-the-loop” is still an overriding principle in AI ethics being discussed — at least for now.)
So the current thinking remains that “responsible use of AI” means we cannot replace human accountability, critical thinking or decision-making.
Over the past month, I have examined a variety of formal approaches to AI ethics and responsible use of AI by public- and private-sector organizations. One organization’s policy that I reviewed requires staff to sign a new AI addendum to their “acceptable use of technology policy,” which dates back years.
After receiving training on AI and what is and is not allowed using checklists and examples, the user must sign an agreement, acknowledging that:
- AI functions by examining extensive data to identify patterns. Using these patterns, it forecasts probable responses, text or results.
- At the same time, AI does not understand meaning, intent, values or context in the way humans do, nor does it possess judgment or self-awareness. AI tool responses are generated based on data patterns and probabilities rather than certainty or understanding.
- Therefore, the user is responsible for all queries/inputs and also accountable for the outputs, ensuring that results are in compliance with the organization’s ethical policies and all applicable laws.
No doubt, many employees and contractors may be uncomfortable with this level of responsibility for moral judgments when their AI tool is doing 99 percent of the data crunching, problem-solving and workflow.
So what are some of the industry frameworks and workable solutions that can help organizations in this ethical AI use area?
YOUR BLUEPRINT FOR AI: ACTIONABLE FRAMEWORKS FOR 2026
Bridging the gap between high-level regulations (like the EU AI Act) and daily workflows doesn’t have to be daunting. Recent research outlines practical checklists to keep your generative AI and data practices safe, ethical and compliant.
There are two frameworks that I want to highlight that deserve attention:
Responsible Prompting: The CLEAR Framework
The detailed CLEAR Prompting Framework can be found here (an alternative PDF download is available here: Download the preprint/accepted manuscript on ResearchGate.)
To ensure employees interface safely with GenAI, adopt the CLEAR Framework methodology for prompt engineering:
- Concise: Eliminate filler text so the model stays focused.
- Logical: Structure instructions in a clear, chronological sequence.
- Explicit: Define the exact scope, tone and strict output constraints.
- Adaptive: Continuously refine and iterate prompts.
- Reflective: Critically evaluate outputs with a mandatory “human-in-the-loop” review.
Quick Check: Never upload proprietary data or PII. Treat AI outputs purely as preliminary drafts requiring manual fact-checking.
Robust Data Governance: The C2V2 Framework Model
The Lifecycle-Based (C2V2) Governance Framework can be found here.
For teams handling data sets and training models, use the C2V2 life cycle checklist to align with global standards:
- Control: Audit data sources, ensure explicit consent, and minimize data retention.
- Consistency: Actively test data sets for bias and representation gaps.
- Value: Keep workflows human-centric, protecting privacy and rights.
- Veracity: Guard against corrupted or contaminated training inputs.
Tip: Balance data usability with FAIR (Findable, Accessible, Interoperable, Reusable) and CARE (Collective Benefit, Authority, Responsibility, Ethics) principles.
AI ETHICS IN BUSINESS
As I was working on this piece, I came across several related articles that I recommend reading. Here they are with brief excerpts:
The American Journal of BioEthics: AI Ethics 2.0: Why Frontier AI Demands a New Governance Agenda for Healthcare — “Traditional AI systems are narrower in scope: a diagnostic algorithm, a scheduling optimizer, a billing classifier. Frontier AI systems, particularly agentic ones, represent something categorically different. They raise at least four distinct governance issues that challenge our current ethical and regulatory frameworks.
“Dynamism: The risks of a frontier AI system are not static. These systems update, drift, and develop emergent behaviors. A system deemed low-risk at the point of deployment does not remain low-risk by virtue of that initial classification alone.
Autonomy: As AI systems carry out longer chains of actions with less human oversight, the potential for unintended consequences grows, partly because the system may begin operating well beyond its original scope.
Interaction: When multiple AI systems interact, coordinating tasks, sharing data, triggering each other’s actions, they can produce systemic risks that no single system would generate on its own. Governance must therefore address the ecosystem as a whole, not just individual tools in isolation.
Context-dependence: Risk emerges from the interaction between a system’s capabilities and its deployment context, user population, level of autonomy, and the reversibility of its decisions. Identical systems deployed in different settings may call for fundamentally different governance approaches.”
UNESCO: Recommendation on the Ethics of Artificial Intelligence – AI ethics for the people
“Core principles: A human rights approach to AI:
1. Proportionality and do no harm: The use of AI systems must not go beyond what is necessary to achieve a legitimate aim. Risk assessment should be used to prevent harms which may result from such uses.
2. Safety and security: Unwanted harms (safety risks) as well as vulnerabilities to attack (security risks) should be avoided and addressed by AI actors.
3. Right to privacy and data protection: Privacy must be protected and promoted throughout the AI lifecycle. Adequate data protection frameworks should also be established.
4. Multi-stakeholder and adaptive governance and collaboration: International law and national sovereignty must be respected in the use of data. Additionally, participation of diverse stakeholders is necessary for inclusive approaches to AI governance.
5. Responsibility and accountability: AI systems should be auditable and traceable. There should be oversight, impact assessment, audit and due diligence mechanisms in place to avoid conflicts with human rights norms and threats to environmental wellbeing.
6. Transparency and explainability: The ethical deployment of AI systems depends on their transparency and explainability (T&E). The level of T&E should be appropriate to the context, as there may be tensions between T&E and other principles such as privacy, safety and security.
7. Human oversight and determination: Member States should ensure that AI systems do not displace ultimate human responsibility and accountability.
8. Sustainability: AI technologies should be assessed against their impacts on sustainability, understood as a set of constantly evolving goals including those set out in the UN’s Sustainable Development Goals.
9. Awareness and literacy: Public understanding of AI and data should be promoted through open and accessible education, civic engagement, digital skills and AI ethics training, media and information literacy.
10. Fairness and non-discrimination: AI actors should promote social justice, fairness, and non-discrimination while taking an inclusive approach to ensure AI’s benefits are accessible to all.”
Another fascinating piece comes from Anthropic’s view on jailbreaks and related issues (using a model in ways that violate guardrails):
“We are therefore partnering with Amazon, Microsoft, Google and other Glasswing partners to draft a consensus framework for assessing the severity of AI jailbreaks and how AI developers should respond to them. We invite other industry partners and model providers to join us in this effort.
“Our current proposal is to score a given jailbreak on the four different criteria below. The first two describe what the jailbreak provides to the attacker; the latter two describe how quickly the jailbreak can become a real-world problem:
- Capability gain. How far beyond existing tools does the jailbreak take the user? If existing widely available tools (including other, weaker AI models) can reach the same capability as the jailbroken model, the score here will be low; if the jailbreak unblocks model capabilities that can significantly accelerate even domain experts, the score will be high.
- Breadth of capability gain. For how many distinct offensive tasks does the same jailbreak technique work? Cases where the jailbreak only allows the model to pursue narrow targets will score low; cases where the same jailbreak technique works for multiple different targets or techniques will score high.
- Ease of weaponization. How much human effort does it take to turn the jailbreak into an attack? Where the jailbreak involves a great deal of skilled prompting and many retries, the score will be low; where the jailbreak works on a single prompt or on the first or second try, the score will be high.
- Discoverability. How easy is it for someone to obtain the technique? If it requires specialist knowledge it will score low; if it is already widely known and available online it will score high.”
And one more thought-provoking related article to consider from Wharton called Why ‘AI Ethics’ Is a Misnomer — and Why That Matters for Business:
“In corporate life, ethics has become compliance: a set of rules, reviews, and risk categories designed to keep organizations out of trouble. That is a legitimate function. It is also a significant narrowing of what ethics actually asks.
“Doing right by people predates machine learning by several millennia. What AI introduces is a new operating environment for an old question: What do we owe one another? Which decisions deserve human deliberation, and which can be delegated to pattern-matching at scale? Ethics is not a new problem that emerged when large language models arrived. It is the oldest problem of humanity, now running inside recommendation engines, clinical triage systems, automated procurement, personalized learning platforms, and intimate consumer chatbots.
“‘AI ethics’ will remain the phrase in circulation. The work it points toward is larger than the phrase suggests. The question for every executive, board member, and strategy team is specific: What kind of natural intelligence (NI) is being cultivated by our AI systems? This question belongs at the beginning of any AI strategy. It belongs in procurement criteria, in design briefs, in performance frameworks, and in the conversations that happen before any contract is signed.
“The answer shapes how we look at two subsequent interrogations: Are our people becoming more capable, more discerning, and more accountable with these tools — or less? Who are we becoming, as an organization, through sustained contact with the systems we are building and buying?”
FINAL THOUGHTS
No doubt, this is a complex topic about which many books have already been written, and many more will be written in the future. Part 2 of this series is intended to provide some guidance and possible frameworks to build solutions.
In summary, here are five core questions to evaluate your organization’s AI ethics and prompting practices:
- How do we ensure human accountability remains central to all AI-driven organizational decisions?
- Are training and policies updated and do they mandate the CLEAR framework (or another framework) for safe, logical and reflective prompting?
- What safeguards prevent staff from inputting proprietary data or PII into AI prompts?
- How do we actively audit our AI tools to prevent algorithmic bias and discrimination?
- Are we treating all AI outputs strictly as preliminary drafts requiring human verification?
Next time, in Part 3 of this series, we’ll be getting a bit more personal and closer to home on AI ethics. Come back for virtual integrity with AI.
