AI leaders doubt self-regulation as online trust gets harder

In short: Many people working in AI do not believe companies can police themselves, and they worry it is becoming harder to know what is real online.

What's going on

Rana Foroohar of the Financial Times reports from a closed-door AI conference in California with senior leaders from AI companies, their customers, and non-profit groups. The meeting came soon after public debate about “rogue AI attacks”, and after President Donald Trump said the answer was “tremendous self-regulation” by AI companies.

Foroohar asked the room if they thought self-regulation would work. She writes that nobody raised a hand. That suggests a gap between political messages and what many people inside the AI industry think is realistic.

In a separate session focused on cyber security, attendees said the technology is moving very fast and that it is getting difficult to tell what is human and what is AI online. One participant cited a striking figure, 108 non-human actors for every human online. “Non-human actors” can include bots (automated accounts that behave like users), and they can make it harder to verify who you are talking to.

Foroohar also describes businesses becoming more wary of each other. Some clients are asking for more customized security, because they do not trust other firms to have protected their systems. She argues this could bring political-style “tribalism” into business, with companies putting up their own “drawbridges” and sharing less.

What to watch

A key question is whether governments can create workable rules for AI when the tools are widely available, and not controlled by a small number of state agencies. The debate is also tied to competition with China, including cheaper “open-source” AI tools (software whose code is shared publicly), which some worry may be harder to secure.

Source: Financial Times