ABBYY Study Finds Business Leaders Disagree on Who Has Ultimate Responsibility When AI Goes Wrong

• 60% of US businesses admit that AI is adapting faster than they can govern it
• Data quality is top reason for not achieving greater ROI
• 70% say data governance requirements slow AI deployment

New global research commissioned by ABBYY reveals that companies are investing heavily in AI, but business leaders are divided over who should ultimately be responsible when AI systems produce harmful or incorrect outputs.

For ABBYY’s 2026 State of Intelligent Automation report, entitled Who Answers for AI: The Governance Gap, Opinium Research surveyed 1,200 senior managers at businesses with 100+ employees in the US, UK, France, Germany, Australia and Singapore.

It found that nearly a third (31%) of business leaders at businesses using AI believe responsibility for harmful or false outputs should be shared between the organization and the vendor. Meanwhile a quarter (26%) believe the organization using the AI should bear primary responsibility, and a fifth (20%) believe it lies fully with the AI vendor. A further 12% say the end user should be responsible, and 10% believe regulators should take the lead.

Lack of clarity around AI governance and responsibility reflects wider AI concerns

Just over a quarter (26%) of business leaders using AI in their organization say they completely trust AI systems to produce accurate outputs, and the same proportion (26%) completely trust AI to explain how important decisions were reached or to protect confidential information. Confidence falls further when it comes to risk, with only 22% saying they completely trust AI systems to operate without introducing unacceptable risks.

These concerns are having a tangible impact on the pace of AI adoption. Over a third (34%) say ethical concerns in their organization considerably slow down deployment of AI initiatives, highlighting the challenge businesses face in balancing innovation with responsible adoption.

Data governance needed to scale AI adoption

Data governance — the controls that ensure data is accurate, secure and compliant — is a major barrier to scaling AI, with 70% of business leaders surveyed saying data governance requirements slow AI deployment.

This could be impacting ROI, with less than half (47%) of global business leaders at organizations using AI saying their AI initiatives have exceeded expectations, although this rises to 53% among US companies.

Concerns around data quality are the biggest barrier to achieving greater ROI from AI, with a fifth (20%) of business leaders citing it as their top challenge. This is followed by integration with existing systems (15%) and implementation costs (13%).

US businesses report stronger AI governance capabilities

The research also showed US businesses report greater confidence in their AI governance capabilities than their global counterparts. Nine in ten (90%) of US organizations say they have adequate systems for auditing AI for compliance, fairness and transparency, compared with 69% in Germany and 76% in both France and Australia. Similarly, 91% of US businesses say data governance and ownership clarity are adequate, compared with 79% globally.

Despite this greater confidence in governance, the US still faces challenges in keeping pace with AI. Six in ten (60%) of US businesses say AI is adapting faster than they can govern it, (versus 54% globally) suggesting that governance maturity does not eliminate the challenge of keeping oversight aligned with the speed of innovation.

Challenges impacting AI ROI

Part of the challenge may be a lack of shared understanding within organizations. Knowledge of who is responsible for AI decreases the more junior employees are. Business leaders at organizations using AI say 89% of senior leadership knows who is responsible for developing, implementing, and managing AI across their organizations, 84% say middle management knows, and 65% say the same for junior employees – although in the US that figure rises to 79%.

AI governance emerges as a critical factor in successful adoption

The findings suggest that effective governance can play a key role in helping organizations realize greater value from AI. 85% of US businesses have an AI governance framework in place and 83% say it has made AI initiatives more successful.

However, governance maturity varies. Only half (50%) of global organizations have an AI incident reporting process, while 46% have an AI incident response plan. Just 35% have an AI kill switch and 35% have AI rollback capabilities, although both figures rise to 41% in the US. Just 4% globally say their organization has no AI incident management capabilities.

More organizations are also planning to strengthen their governance approach, with 15% of business leaders at organizations using AI saying their organization does not currently have official AI governance in place but plans to develop one within the next 12 months.

Those with AI governance frameworks in place are already seeing practical benefits – especially in the US. 63% of US respondents say it makes developing AI systems easier, while 61% said it makes scaling past pilot stage easier, and 66% said it makes adoption by employees easier.

Human oversight important to AI governance

Human oversight remains important, with 56% of business leaders globally saying humans review all important AI decisions before action is taken. This rises to 60% in the US, 61% in Australia, while falling to 49% in Singapore, showing continued variation in how organizations approach human involvement in AI decision-making.