From 80,000 to 1.5 Million Users: US Fears Reliance on AI May Turn into a Threat
Within a single year, the US military will make more progress than in the previous five years. This statement from a high-ranking Pentagon official comes amidst new concerns: artificial intelligence is now approaching capabilities that can be utilised as cyber weapons.
It is no longer merely about chatbots. The United States Department of Defense has begun warning that the latest generation of AI models possesses the ability to discover security loopholes, exploit software vulnerabilities, and potentially be used in large-scale hacking operations. The pressing question remains: who will oversee this increasingly powerful technology?
These concerns were expressed by Emil Michael, the US Deputy Secretary of Defense for Research and Engineering. During the Building America Summit organised by The Washington Post a few days ago, Michael stated that AI companies bear a significant responsibility to ensure their technology does not transform into a threat to society.
“These companies have a responsibility to ensure that what they call the weaponisable capabilities of these models are handled with care and prudence,” he remarked.
This warning follows the White House issuing a new executive order regarding artificial intelligence. Under this policy, the US government is establishing an AI cybersecurity information centre, allowing technology companies to voluntarily submit their models for testing by the Pentagon before they are released to the public. The reason for this government intervention is that the threat is considered real.
Policymakers in Washington are increasingly worried that AI with advanced cybersecurity capabilities could be exploited by adversary nations, criminal groups, or hacking actors to attack power grids, hospitals, government institutions, and other critical infrastructure. The spotlight is particularly on the emergence of new models with increasingly sophisticated offensive capabilities. One such example is Mythos, the latest AI model from Anthropic. The company has acknowledged that the model is capable of identifying and exploiting weaknesses in software. While proponents of the technology argue these capabilities are useful for strengthening digital security, critics fear the same abilities could be turned into dangerous tools for attack.
Source: jawawa.id (https://jawawa.id/newsitem/from-80-000-to-1-5-million-users-us-fears-reliance-on-ai-may-turn-into-a-threat-1780872532)
Originally published by Republika (https://news.republika.co.id/berita/tg9w3r451/80-ribu-meledak-jadi-15-juta-user-as-khawatir-ai-yang-mereka-andalkan-malah-berbalik-jadi-ancaman), translated from Indonesian