November 21, 2026 USD 44.93EUR 49.28PLN 11.06
Новини Марганець — коротко й по суті
LIVE
Over Half of Damaged Healthcare Facilities Restored in Chernihiv Region What to do if your smartphone gets wet: 4 common mistakes to avoid EU to Provide €1 Billion for Ukraine's Defense: Key Details Financial Imbalance: EU Spending on Russian Energy Exceeds Aid to Ukraine Another attack on Zaporizhzhia: kindergarten and residential building damaged Ukraine blocks Russia's attempts to legitimize occupation at COP31 climate summit Missile strike on Kryvyi Rih: civilians killed, dozens injured Zelenskyy reaffirms that the Crimean Bridge is a legitimate target Nighttime attack on Zaporizhzhia: kindergarten and residential building damaged Massive attack on the energy system: Rivne region partially blacked out Over Half of Damaged Healthcare Facilities Restored in Chernihiv Region What to do if your smartphone gets wet: 4 common mistakes to avoid EU to Provide €1 Billion for Ukraine's Defense: Key Details Financial Imbalance: EU Spending on Russian Energy Exceeds Aid to Ukraine Another attack on Zaporizhzhia: kindergarten and residential building damaged Ukraine blocks Russia's attempts to legitimize occupation at COP31 climate summit Missile strike on Kryvyi Rih: civilians killed, dozens injured Zelenskyy reaffirms that the Crimean Bridge is a legitimate target Nighttime attack on Zaporizhzhia: kindergarten and residential building damaged Massive attack on the energy system: Rivne region partially blacked out

Artificial Intelligence under scrutiny: Chatbots vulnerable to manipulation

💡 Tech 1 min read 👁 0

A US Government Accountability Office report reveals that popular AI models can provide instructions for planning attacks after a series of targeted prompts.

Artificial Intelligence under scrutiny: Chatbots vulnerable to manipulation💡Tech

Photo: RBC-Ukraine

Key points

  • Four leading AI models failed to maintain safety protocols
  • Detailed attack plans were generated in under 12 hours
  • U.S. Congress announced hearings on AI usage risks

Vulnerability of modern technology

A large-scale study conducted over nearly a year has exposed significant security gaps in leading AI chatbots, including ChatGPT, Claude, Gemini, and Grok. Experts from the U.S. Government Accountability Office (GAO) found that users without specialized technical knowledge could manipulate artificial intelligence into generating detailed plans for dangerous activities.

Mechanism of bypassing restrictions

During testing, researchers simulated the behavior of average users. While direct requests for harmful instructions were blocked, a sequential dialogue allowed researchers to bypass these safeguards in less than 12 hours. Consequently, the systems provided specific recommendations regarding target selection, logistics, and methods of causing harm, including attacks on critical infrastructure.

Senator Maggie Hassan emphasized the critical importance of this issue for national security, urging Congress to take immediate action. Tech developers have stated they are continuing to refine their platforms and strengthen safety mechanisms to minimize the risks of AI being used for criminal purposes.

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.