10
Loading video...
By L. Stephan Vincze
Founder and CEO
0
Share
Newsweek is a Trust Project member
See more of our trusted coverage when you search.Prefer Newsweek on Googleto see more of our trusted coverage when you search.
Artificial intelligence (AI) is already influencing healthcare, scientific research, and business decisions that shape people’s lives. Yet organizations are racing to deploy it without necessarily establishing who is accountable when it fails.
That is a leadership failure. Patients, consumers, employees, and investors all have a stake in whether companies can prove their AI is being used responsibly. They should demand that proof.
The debate about AI is too often trapped between catastrophic predictions and promises of extraordinary productivity.
Read More on Leadership Insights
Both distract from the immediate problem: can an organization identify where AI is being used, establish who owns each application, and produce evidence that its outputs are checked?
If it cannot, its assurances about responsible AI are public relations, not governance.
...
The evidence reinforces the urgency. In its 2026 AI Index Report, Stanford University’s Human-Centered AI Institute found that documented AI incidents recorded by the AI Incident Database rose from 233 in 2024 to 362 in 2025.
The report also found that responsible AI benchmarking and reporting are struggling to keep pace with advances in AI capabilities. AI is advancing rapidly, but the systems needed to assess and manage its risks are not keeping up.
This matters far beyond the boardroom. In healthcare and life sciences, unchecked AI could miss an adverse event, distort a clinical finding, or generate an unsupported claim about a medicine.
In other industries, it could expose private data, mislead customers, or influence decisions about money and access to essential services.
Ordinary people cannot be expected to determine whether a harmful decision originated with a human or a machine. They need institutions that can explain what happened and accept responsibility.
Accountability Theater
I call one of the central dangers moral outsourcing: allowing AI to substitute for accountable human judgment. Arguing that the 'system said so' is not a defense. Neither is blaming a vendor or pointing to a policy that nobody followed.
A human reviewer who lacks the time, training, or authority to challenge an AI-generated answer is not meaningful oversight. It is accountability theater.
There is a better path. Properly governed AI can accelerate research, improve services, and free people from routine work. But organizations must establish three non-negotiable controls.
First, inventory every AI tool in use. Identify approved systems, third-party applications, and tools employees may have adopted without authorization. Invisible systems cannot be governed.
Second, assign a named, accountable human owner to every tool and use case. That person must understand the system’s purpose, have authority to escalate concerns, and be empowered to suspend unsafe use. Responsibility must exist in practice, not merely on an organizational chart.
Third, verify outputs and document the checks. Record what was reviewed, who reviewed it, what evidence was considered, and how problems were resolved.
The intensity of scrutiny must reflect the potential consequences. An AI-generated meeting summary does not carry the same risk as an output influencing patient safety or scientific conclusions.
Effective Governance
These controls must work together. An inventory without ownership is a list. Ownership without verification is an assumption. Verification without records is difficult to demonstrate when a board, regulator, or injured customer demands answers.
Nor can executives delegate this responsibility entirely to IT departments or software vendors. Technical teams can explain how a system operates, but leaders remain accountable for how their organizations use it.
Effective governance must evolve as models, data, and applications change. A policy filed away on a shared drive cannot detect a deteriorating system or prevent an employee from relying on a dangerous output.
The choice is stark. With effective oversight, organizations can innovate with greater confidence, catch failures earlier, protect the people who depend on them, and earn public trust.
Without it, unchecked systems and blurred responsibilities create the conditions for preventable harm. When the inevitable questions arrive, leaders may discover they cannot establish who made the decision, what checks were performed, or whether anyone was watching.
Boards and executives must commission an organization-wide AI integrity assessment now. Map every tool, assign accountable owners, test outputs according to risk, document safeguards, and close identified gaps. Repeat the assessment on a regular basis.
Before your organization expands its use of AI, demand evidence that someone is responsible, outputs are checked, and safeguards work. If you cannot prove you are in control, you have no business claiming that your AI is responsible.
L. Stephan Vincze is an attorney, founder and CEO of TRESTLE Compliance, which provides AI governance and compliance services to healthcare and life sciences organisations. All opinions are the author's own.
View 0 Comments
Submit Correction
View Editorial & AI Guidelines