TOP AI RISKS & THREATS IN 2026 - PART II
This is Part II of the AI risks & threats landscape in 2026
Earlier this month, we took you beyond the AI hype by outlining the top AI risks and threats in 2026 and beyond that we should all be concerned about. We shared our observation that the most dangerous AI risks & threats are part of the design and often intentional. This is part II in our series, outlining the last 5 out of 10 AI risks & threats, that WAIE+ is currently tracking.
6. EXPOSURE OF SENSITIVE USER CHATS
Fast Company reported that thousands of ChatGPT conversations were found in Google search results.
Leakage of users’ personal details in the public sphere can make it possible to identify the user and expose them to serious risks.
7. MALICIOUS CODE IN AI TOOLS
The number of freely available AI tools with malicious code is growing.
In 2024, a Disney employee downloaded an AI tool from GitHub containing malware that granted hackers access to the company’s confidential data and customer information.
8. EMOTIONAL MANIPULATION
The parents of multiple teenagers have filed wrongful death suits against OpenAI alleging that ChatGPT advised their child on methods to kill himself.
AI systems are optimized to use the most effective way to achieve their objective and help users even if the goal is self-harm.
Beyond the AI Hype by WAIE+ is a reader-supported publication. To receive new posts and support our work, consider becoming a free or paid subscriber.
9. DECEPTION & FRAUD
Criminals are using AI deepfakes to dupe people into sending money to fraudulent accounts.
Extortion scams, often as phone calls that sound like a family member in trouble, exploit victims to get payment or sensitive information.
10. JOB LOSSES
Jobs displacement is not typically included in mainstream discourse of AI risks and threats. However, it is a critical issue that needs to be addressed urgently.
Rapid expansion of AI automation and use of original content as training data is eroding creators’ livelihoods and eliminating entry-level jobs.
