November 22, 2026 USD 44.93EUR 49.28PLN 11.06
Новини Нікополь — коротко й по суті
LIVE
Unique Polish Village of Suloszowa Amazes with a Single 9-Kilometer-Long Street Odesa Schools Switch to Remote Learning Due to Utility Disruptions Strike on Dymer and Massive Shelling Across Regions: People Trapped Under Rubble Utility providers step up pressure on debtors: water cutoffs announced in neighboring community The Philosophy of Reconstruction: Why Everyday Repairs Are Becoming the Foundation of Survival for Frontline Cities Elina Svitolina and Marta Kostyuk Nominated for Prestigious WTA Awards for the 2026 Season Russian forces target energy infrastructure in Zhytomyr region four times in a day Robert Fico makes harsh statement regarding new €45 billion EU loan for Ukraine High-Rise Damaged in Kyiv Region: 14 People Unaccounted For After Drone Attack Nikopol Youth Club Hosts Special Event for World Children's Day Unique Polish Village of Suloszowa Amazes with a Single 9-Kilometer-Long Street Odesa Schools Switch to Remote Learning Due to Utility Disruptions Strike on Dymer and Massive Shelling Across Regions: People Trapped Under Rubble Utility providers step up pressure on debtors: water cutoffs announced in neighboring community The Philosophy of Reconstruction: Why Everyday Repairs Are Becoming the Foundation of Survival for Frontline Cities Elina Svitolina and Marta Kostyuk Nominated for Prestigious WTA Awards for the 2026 Season Russian forces target energy infrastructure in Zhytomyr region four times in a day Robert Fico makes harsh statement regarding new €45 billion EU loan for Ukraine High-Rise Damaged in Kyiv Region: 14 People Unaccounted For After Drone Attack Nikopol Youth Club Hosts Special Event for World Children's Day

GAO report: popular AI chatbots vulnerable to generating attack plans

💡 Tech 2 min read 👁 1

US auditors have found that leading artificial intelligence models can generate dangerous instructions after being manipulated by users.

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

Photo: RBC-Ukraine

Key points

  • GAO identified vulnerabilities in four leading chatbots
  • Users were able to generate attack plans within 12 hours
  • US Senate to hold hearings on AI safety

Vulnerability of artificial intelligence systems

A comprehensive year-long study by the US Government Accountability Office (GAO) has uncovered critical security flaws in leading chatbots such as ChatGPT, Claude, Gemini, and Grok. Experts demonstrated that even users without specialized technical knowledge can bypass built-in safety restrictions by using a series of probing, follow-up prompts.

The report notes that in less than 12 hours of interaction with the AI, researchers were able to generate detailed attack scenarios. These plans included target selection, timelines, logistics, and methods to evade law enforcement. Independent experts described the gathered material as realistic and potentially dangerous.

Risks to critical infrastructure

Certain platforms demonstrated the ability to provide instructions on constructing explosive devices or planning sabotage against energy facilities. Specifically, some models suggested using drones to damage substations, accompanying their advice with visual simulations of the potential impact.

«A bad actor could use a common AI chatbot and in less than 12 hours have a roadmap for a lethal attack,» said Senator Maggie Hassan.

Technology developers, including OpenAI, Google, and Anthropic, have stated their commitment to strengthening safety protocols. Meanwhile, the US Congress has already announced hearings to address the risks posed by uncontrolled access to such tools.

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.