A study led by researchers from MIT Future Tech and the University of Queensland, published in summer, identified 24 risks associated with artificial intelligence, with 18 posing at least a 10% chance of catastrophic consequences. The research, titled “Prioritizace rizik od umělé inteligence,” involved 272 international AI experts who evaluated risks based on probability and severity. The study found that under a “status quo” scenario, 18 of the 24 risks would have a 10% or higher likelihood of causing catastrophic outcomes by 2030, defined as over a million deaths, financial losses exceeding $100 billion, or comparable non-material damage.
The researchers used a multi-round evaluation process to identify areas of consensus and disagreement, then compared results under two scenarios: maintaining the current approach to AI development and implementing targeted risk mitigation. While the latter reduced the number of high-risk scenarios, five risks remained with at least a 10% chance of catastrophic consequences. These included AI systems with dangerous capabilities, competitive dynamics in AI development, cyberattacks, weaponization, and misinformation. Peter Slatery, a MIT Future Tech researcher and co-author, noted that AI could enable activities like cyberattacks and weapon development “much more easily” than before.
Key Risks Identified
The study highlighted specific risks, including AI systems capable of causing significant harm through manipulation, cyberattacks, or autonomous development. It also cited concerns about competitive pressures leading to reduced safety measures in AI development. Cybersecurity threats and the creation of misleading content were emphasized as growing dangers. The report noted that the information sector, national security, and finance were the most vulnerable to these risks, with recent events validating their validity.
Among the risks identified, unequal power distribution emerged as a critical concern. If the most advanced AI systems remain concentrated in the hands of a few leading companies or governments, they could consolidate political and economic power, disadvantaging broader segments of society. This imbalance was described as “extraordinarily unjust” in the study.
Methodology and Expert Involvement
The research team included experts from MIT Future Tech and the University of Queensland, who collaborated to prioritize risks through iterative expert feedback. The study’s methodology involved analyzing risks under two distinct scenarios, with the “status quo” approach yielding the highest number of critical risks. While targeted mitigation reduced some threats, the findings suggest that even with interventions, five risks would still carry significant potential for catastrophic outcomes by 2030.
Interpol’s Warning on AI and Cybercrime
Bjorn R. Watne, chief information officer (CISO) of Interpol, warned that artificial intelligence would not revolutionize cybercrime but would instead enhance existing techniques such as phishing and fraud. “This is a process of evolution, not a revolution,” he said at Tech Week Singapore on October 1. Watne emphasized that AI increases the “speed and scope” of criminal activities, enabling fraud rings to target more victims simultaneously. Improved translation technologies and digital identity systems also make it harder to distinguish fake interactions from real ones.
Watne urged companies to identify their “most critical assets” rather than trying to protect all systems at once. “Organizations should focus on assets vital to their operations and assess who might want to steal or damage them,” he said. He also highlighted the growing risk of AI agents acting autonomously on behalf of users, which could lead to real-world harm if errors occur. For example, self-driving cars integrating AI raise concerns about overreliance on technology, as users tend to trust it excessively.
Interpol also noted a gap between executive leadership and cybersecurity strategies in many organizations. Despite rising cyber threats, this issue remains underaddressed in corporate boards.
Upcoming IT Conferences Highlight AI’s Impact
The 14th edition of the HackerFest cybersecurity conference, scheduled for October 8, 2026, will feature presentations on advanced attack techniques. Lubomír Ošmera will demonstrate methods effective in Windows Server 2025 environments without NTLM. William Ischanoe will analyze the Windows Subsystem for Linux 2 (WSL2) from an attacker’s perspective, while Roman Kümmel will discuss network service audits. Jan Marek will explore memory internals in Windows.
Registration for in-person attendance at HackerFest 2026 costs 5,400 Kč, with online access available for 3,500 Kč (excluding VAT). Participants will also receive a 10-day archive of lectures. The JavaDays 2026 conference, held November 25–26, 2026, will focus on AI’s influence on software development. Jan Novotný will present an open-source MCP server enabling AI agents to access Java debuggers, while Petr Adámek of Oracle will address trustworthy code in the AI era.
Ticket prices for JavaDays 2026 are 7,900 Kč for in-person attendance and 4,500 Kč for online participation. A 14-day archive of lectures is included.
Worth a look