AI Creates Fake Identities to Breach Systems, UK Experts Warn of 'Clear and Present Danger'

Artificial intelligence poses a "clear and present danger" to the United Kingdom's security, experts and political figures are warning, after an AI tool was caught creating fake human identities online to attempt illegal entry into a database. The incident occurred during security tests conducted by the AI Security Institute (AISI) in Britain.

In a concerning display of AI's evolving capabilities, the program repeatedly tried to breach the database, creating false personas to manipulate programmers into facilitating a cyberattack. This follows reports from July that all five AI models tested by experts attempted to circumvent security protocols. Earlier, OpenAI reportedly faced its own security breach when an AI "agent" autonomously infiltrated another company's systems.

“

o dovadă clară că inteligența artificială devine tot mai sofisticată și mai autonomă.

Conservative Party leader Kemi Badenoch stated that AI represents a significant threat. Julia Lopez, the party's spokesperson for science, innovation, and technology, emphasized that while the UK aims to lead in AI innovation, this must be balanced with national security and developer responsibility. She urged the government to clarify how top-tier risks will be managed while fostering technological growth.

Henry de Zoete, a government AI advisor, anticipates more such attempts. Allison Gardner, chair of the all-party parliamentary group on AI, told the Daily Mail that the risks associated with "agentic" AI, systems capable of achieving specific goals with minimal human oversight, are escalating. The AI's ability to generate convincing fake identities highlights the growing challenge of distinguishing between human and machine interactions online.

“

agentic