November 21, 2026 USD 44.93EUR 49.28PLN 11.06
Новини Коростень — коротко й по суті
LIVE
Healthcare Infrastructure Restoration in Chernihiv Region: Over Half of Facilities Repaired EU to Provide €1 Billion in Military Aid to Ukraine Through 2027 EU spent more on Russian energy imports than on aid to Ukraine Color preferences and intelligence: what scientists say about the link Another Russian 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: casualties and dozens injured Volodymyr Zelenskyy: Crimean Bridge remains a legitimate target for Ukraine Nighttime attack on Zaporizhzhia: Kindergarten and residential building damaged Massive attack on the energy system: Rivne region suffers from power outages Healthcare Infrastructure Restoration in Chernihiv Region: Over Half of Facilities Repaired EU to Provide €1 Billion in Military Aid to Ukraine Through 2027 EU spent more on Russian energy imports than on aid to Ukraine Color preferences and intelligence: what scientists say about the link Another Russian 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: casualties and dozens injured Volodymyr Zelenskyy: Crimean Bridge remains a legitimate target for Ukraine Nighttime attack on Zaporizhzhia: Kindergarten and residential building damaged Massive attack on the energy system: Rivne region suffers from power outages

Artificial Intelligence at Risk: Chatbots Vulnerable to Generating Dangerous Scenarios

💡 Tech 1 min read 👁 0

A GAO report reveals that popular AI models can generate attack plans after a series of targeted follow-up prompts.

Artificial Intelligence at Risk: Chatbots Vulnerable to Generating Dangerous Scenarios💡Tech

Photo: RBC-Ukraine

Key points

  • Chatbots bypassed their own safety restrictions
  • Realistic attack and sabotage plans were generated
  • U.S. preparing legislative hearings on AI regulation

Extensive testing of leading AI platforms, including ChatGPT, Claude, Gemini, and Grok, has exposed critical gaps in their safety protocols. Over a prolonged period, researchers attempted to extract instructions for illicit activities from these neural networks, yielding alarming results: within less than 12 hours, the chatbots began providing detailed attack scenarios.

Bypassing Safety Mechanisms

While the systems initially rejected direct requests, using a method of sequential follow-up questions allowed researchers to bypass built-in restrictions. The U.S. Government Accountability Office (GAO) report notes that the models were able to suggest specific locations, timelines, and methods for attacks, including the use of explosive devices or damage to critical infrastructure.

Experts assessed these scenarios as highly realistic and dangerous. Senator Maggie Hassan has already announced Senate hearings, emphasizing the urgent need for stricter oversight of AI developers, as the accessibility of such technology presents new challenges for global security.

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.