BMO Says AI Safety Concerns Unlikely to Slow Pace of Frontier Development

Leading artificial intelligence labs are unlikely to slow their pace of frontier model development, despite expressing genuine and valid concerns about the technology's safety, BMO said in a research note on Friday.

The bank's analysis suggests a disconnect between the stated fears of AI leaders and the continued release of increasingly powerful models from developers like Anthropic and OpenAI.

Development Pace vs. Safety Fears

Calls for a coordinated slowdown in AI development, most notably from Anthropic CEO Dario Amodei, are rooted in the need to align advanced systems with human interests before potential problems become irreversible, BMO noted. Several AI executives have supported this proposal, and the topic was reportedly discussed during a recent U.S.-China summit.

However, BMO questioned the commitment of major developers to self-regulation, pointing to their ongoing product launches. The note also addressed security, stating that while recent AI-related hacking incidents have been "largely benign," the risk of bad actors exploiting open-access models is growing as users find ways to override built-in safeguards.

Regulatory and Market Implications

Ad

BMO expects both the U.S. and China to eventually tighten AI regulation, with a focus on open-source and open-access models rather than closed systems. Such rules would likely require government approval before the broad public release of these models.

This regulatory shift would benefit closed-source frontier labs, according to the note, by limiting competition and strengthening their pricing power. BMO forecasts that AI token prices will continue to fall, boosting user adoption, before bottoming out around 2030. After this point, closed-source providers are expected to raise prices as users become more dependent on their platforms.

Economic and Sector Outlook

Looking at the broader market, BMO offered several key takeaways:

• AI Hardware: The bank remains bullish on the hardware sector, citing a growing scarcity of computing power as demand for AI capabilities outstrips earlier expectations.
• Meta Platforms (META): While the company has improved its AI positioning, it still lags behind frontier leaders like Anthropic and OpenAI.
• Job Displacement: BMO projects that white-collar job displacement will become meaningful by 2028. Blue-collar roles are expected to be impacted later, in the 2030s, a delay the bank attributes to Moravec's Paradox, which posits that tasks requiring physical dexterity are more difficult for AI to automate than cognitive ones.