AI development has major security, privacy and ethical blind spots

Security, privacy and ethics are low-priority issues for developers when modeling their machine learning solutions, according to O’Reilly.

Major issues

Security is the most serious blind spot. Nearly three-quarters (73 per cent) of respondents indicated they don’t check for security vulnerabilities during model building. More than half (59 per cent) of organizations also don’t consider fairness, bias or ethical issues during ML development.

Privacy is similarly neglected, with only 35 per cent checking for issues during model building and deployment.

Risks during development

Instead, the majority of developmental resources are focused on ensuring artificial intelligence projects are accurate and successful. The majority (55 per cent) of developers mitigate against unexpected outcomes or predictions, but this still leaves a large number who don’t. Furthermore, 16 per cent of respondents don’t check for any risks at all during development.

This lack of due diligence is likely due to numerous internal challenges and factors, but the greatest roadblock hindering progress is cultural resistance, as indicated by 23 per cent of respondents.

AI maturity

The research also shows 19 per cent of organizations struggle to adopt AI due to a lack of data and data quality issues, as well as the absence of necessary skills for development. The most chronic skills shortages by far were centered around ML modeling and data science (57 per cent). To make progress in the areas of security, privacy and ethics, organizations urgently need to address these talent shortages.

“AI maturity and usage has grown exponentially in the last year. However, considerable hurdles remain that keep it from reaching critical mass,” said Ben Lorica, chief data scientist, O’Reilly.

“As AI and ML become increasingly automated, it’s paramount organizations invest the necessary time and resources to get security and ethics right. To do this, enterprises need the right talent and the best data. Closing the skills gap and taking another look at data quality should be their top priorities in the coming year.”

Other key findings

• The overwhelming majority of organizations (81 per cent) have started down the route of AI adoption. Most are in the evaluation or proof of concept stage (54 per cent), while 27 per cent have revenue-bearing AI projects in production.
• A significant minority (19 per cent) of companies have not started any AI projects.
• Machine learning has emerged as the most popular form of AI used by enterprises. Nearly two-thirds (63 per cent) use supervised learning solutions while 55 per cent are using deep learning technology. Model-based methods are used by almost half (48 per cent) of respondents.
• AI is most likely to be used in research and development (R&D) departments (50 per cent), customer service (34 per cent) and IT (33 per cent). Legal functions have seen the least innovation, with only 5 per cent making use of AI technologies.
• TensorFlow (55 per cent) and scikit-learn (48 per cent) are the most popular AI tools in use today.