AI math progress raises new concerns about online encryption security

In short: As AI gets better at advanced maths, experts and regulators are rechecking whether today’s encryption will stay safe.

What's going on

Encryption is the lock that protects digital information, like online banking logins, private messages, and company databases. For years, many security plans assumed the biggest future threat would come from quantum computers, which are still seen as years away.

Now some mathematicians and cyber security experts are paying closer attention to AI. They say recent progress suggests AI could help solve certain hard maths problems faster than expected, and those problems are what many encryption methods rely on.

The Financial Times points to a recent dispute involving OpenAI, after it released hundreds of mathematical papers that it said included AI generated solutions and proofs. Some researchers argued this kind of progress could change what counts as “difficult” in maths, which matters because encryption depends on tasks that are easy one way and very hard the other way.

A common example is encryption based on multiplying two huge prime numbers (numbers only divisible by 1 and themselves). Multiplying them is easy for computers, but working backward to find the two original primes is supposed to be extremely hard. Experts are debating whether AI could make that reverse step easier.

What to watch

Regulators are already pushing organizations to prepare. The UK’s Financial Conduct Authority said it is testing a new tool with 50 large institutions to check how ready they are for new risks from AI and quantum computing, and it plans to share guidance next month. Some experts also warn that some newer “post-quantum” encryption options (meant to resist quantum computers) may be less proven and could have weak spots if AI improves quickly.

Source: Financial Times