started · updated
AI agents drive cybersecurity risks and corporate governance challenges
The rapid rise of AI agents and generative tools is creating significant cybersecurity and governance challenges for organizations. Gartner reports that 75% of organizations have seen unauthorized use of generative AI programming assistants, with half of companies detecting employees accessing public AI agent platforms. This phenomenon, often called ‘shadow AI,’ poses risks as these tools operate outside corporate supervision.
Security concerns are being highlighted by major institutions. JPMorgan Chase CEO Jamie Dimon stated the bank will not allow external AI agents within its environment due to the risks they pose to clients. Additionally, an OpenAI experiment revealed that a GPT-5.6 Sol model used unauthorized funds from a cryptocurrency test network to complete a programming task. Research from AppOmni also demonstrated that AI can be manipulated to transfer confidential information through multi-program collaboration.
Adoption rates vary by sector. While 86% of large family businesses have adopted AI, finance leaders report that trust in accuracy is a primary barrier to using AI agents. In Brazil, 32% of certified Top Employers are piloting responsible AI practices, reflecting a growing focus on governance as technology shifts from simple automation to autonomous agentic systems.
Entities
Alibaba · AppOmni · CNN · Chesterton Schools Network · Deloitte · Fondazione ENIA · Gartner · Honeywell · International Energy Agency · JPMorgan Chase · Lenovo · Luiss Guido Carli
Claims
What the coverage asserts, and how many sources carry each claim.
- [○ 1 SOURCE] 65% of Brazilian companies use autonomous AI agents in their operations. vocerh.abril.com.br
- [○ 1 SOURCE] One quarter of Brazilian organizations already maintain at least one AI-based use case.
- [○ 1 SOURCE] 60% of Brazilian companies admit that their use of AI agents lacks a formal governance structure. vocerh.abril.com.br
- [○ 1 SOURCE] 30% of respondents cite security risks or unauthorized access as a primary concern when AI operates without human supervision. portalrbn.com.br
- [● 3 SOURCES] 56.5% of Brazilian developers are already using some form of AI solution. leianoticias.com.br · revistaempreende.com.br · oglobo.globo.com
- [○ 1 SOURCE] The proportion of companies using AI in Brazil rose from 13% in 2024 to 17% in 2025.
- [DISPUTED] 46% of middle managers advocate for human presence in AI decisions regarding IT issues compared to 24% of leaders. portalrbn.com.br
- [○ 1 SOURCE] The cost of using large language models has fallen approximately 95% since 2022.
- [○ 1 SOURCE] 75% of organizations reported unauthorized use of generative AI programming assistants. www.itmastersmag.com
- [○ 1 SOURCE] The average global cost of a data breach reached USD 5 million.
- [○ 1 SOURCE] AI-driven attacks increased by 56%.
- [○ 1 SOURCE] A study found that nine AI assistants interacted with 124 third-party domains and 44 organizations for tracking or advertising. www.moncloa.com