From the Atlantic:
Since late last year, a small but growing chorus of cybersecurity experts has been warning that AI models were pushing the internet toward a sort of cataclysm. Hospitals, banks, utilities, basically anything and everything connected to the web could be hacked.
Over the past month or so, those scenarios have come alive in almost spectacular fashion: Last month, AI was used to find major security flaws in software used by 70 percent of websites, prompting governments to take down their websites as a precaution. A week later, attackers targeted the world’s largest cryptocurrency exchange, Binance, in part by hijacking the company’s internal AI tools, and ultimately drained billions of dollars. This fall alone, AI has been used to hack hundreds of financial institutions in Latin America, not to mention data centers, banks, and government entities around the world. AI has also changed the economics of hacking, such that targeting individual people and small organizations is becoming cheap and easy enough to be worthwhile. And there’s a broader rise in cyberattacks that, though not directly tied to AI, likely leveraged the technology in some way. Last week, for instance, the FBI issued a warning about hacks targeting industrial refrigeration systems following attacks on grocery stores in France, Belgium, Ireland, Sweden, Israel, and other countries.