November 22, 2026 USD 44.93EUR 49.28PLN 11.06
Новини Жмеринка — коротко й по суті
LIVE
Belgium to Completely Halt Air Traffic Due to Nationwide Strike Russian vessel Sparta II still on fire in Gulf of Finland following Ukrainian drone attack NABU Initiates Cases Against SBU Special Forces Over Past Searches From Intensive Care to Her Own Business: How a Vinnytsia Woman Started Making Japanese Sweets After Severe Burns Strike on Dymer, Kyiv Region: Search Operations Underway at Damaged Apartment Building Behind the Scenes of European Tours: How Ukrainian Theaters Unite the Diaspora Famous Singer Miley Cyrus Secretly Marries Musician Max Morando Enemy Attacks Energy Infrastructure in Neighboring Zhytomyr Region Over Half a Million for Partnership: Ukrainian Researchers Eligible for International Grants The Rarest Document on Earth: Who Receives the Exclusive Maltese Order Passports Belgium to Completely Halt Air Traffic Due to Nationwide Strike Russian vessel Sparta II still on fire in Gulf of Finland following Ukrainian drone attack NABU Initiates Cases Against SBU Special Forces Over Past Searches From Intensive Care to Her Own Business: How a Vinnytsia Woman Started Making Japanese Sweets After Severe Burns Strike on Dymer, Kyiv Region: Search Operations Underway at Damaged Apartment Building Behind the Scenes of European Tours: How Ukrainian Theaters Unite the Diaspora Famous Singer Miley Cyrus Secretly Marries Musician Max Morando Enemy Attacks Energy Infrastructure in Neighboring Zhytomyr Region Over Half a Million for Partnership: Ukrainian Researchers Eligible for International Grants The Rarest Document on Earth: Who Receives the Exclusive Maltese Order Passports

GAO report: popular AI chatbots vulnerable to generating dangerous attack plans

💡 Tech 1 min read 👁 0

US auditors found that leading AI models can provide detailed attack scenarios after a series of follow-up questions.

GAO report: popular AI chatbots vulnerable to generating dangerous attack plans💡Tech

Photo: RBC-Ukraine

Key points

  • AI models provided attack plans after prompt manipulation
  • The study lasted nearly a year under GAO oversight
  • Developers pledge to strengthen security measures

A comprehensive study by the U.S. Government Accountability Office (GAO), conducted between 2025 and 2026, has uncovered significant security flaws in leading AI chatbots. Experts tested four popular platforms, including systems from OpenAI, Google, Anthropic, and SpaceXAI, to evaluate their resilience against manipulation.

Bypassing security protocols

While the AI systems initially refused direct requests for terrorism-related plans, researchers successfully bypassed these safeguards. By employing a technique of iterative follow-up questions, they obtained detailed instructions in less than 12 hours. The generated output included attack scenarios involving explosive devices, drones, and firearms, targeting energy infrastructure and crowded public venues.

Senator Maggie Hassan highlighted the danger posed by such accessibility, urging Congress to take immediate action. Technology developers have stated they are continuing to enhance security mechanisms, yet the GAO report emphasizes that risks to national and economic security remain a pressing concern.

Source RBC-Ukraine — rbc.ua/ukr/news/chatgpt-claude-gemini-ta-grok-stvorili-plani-179518… ↗

Contact the newsroom

Got a tip, a proposal or a partnership question? Write to us and we will reply shortly.