And speaking of Defcon, AI security researchers are exploring new methods for reporting flaws in AI models at the DEF CON conference in Las Vegas. Current practices lack standardization, making it difficult for companies to receive and address findings from ethical hackers. The AI Village is collaborating with the Allen Institute for AI to test an open-source model and refine the reporting process for security bugs, aiming to improve how vulnerabilities are disclosed and handled in AI systems.
OpenAI has pledged to provide the U.S. AI Safety Institute early access to its next generative AI model for safety testing, aiming to address concerns about its commitment to AI safety. This follows criticism over the disbandment of a safety team and the company’s recent moves to eliminate non-disparagement clauses and create a safety commission. The agreement coincides with OpenAI’s endorsement of a Senate bill that would establish the Safety Institute, raising concerns about potential regulatory influence.
Speaking of OpenAI, the company has experienced significant leadership changes, with president Greg Brockman taking a leave of absence, vice president Peter Deng leaving, and co-founder John Schulman joining rival Anthropic. These departures follow a trend of top-level exits at OpenAI, leaving only three of the original eleven founders with the company. Schulman cited a desire to focus on AI alignment as his reason for leaving, while Brockman expressed the need for a sabbatical after nine years with OpenAI.
A survey of 300 American tech workers reveals strong support for AI regulation, with 75% advocating for federal and state oversight to address privacy, security, and ethical concerns. The survey highlights the need for updated copyright laws to protect content creators from data misuse by AI companies. While the US has some regulatory efforts, such as an executive order from Biden, comprehensive frameworks like the EU’s AI Act are still in development. Despite regulatory gaps, 88% of respondents trust their employers to use AI responsibly.
Regulation is more popular than one might think, and Open AI recognizes that. That said, there’s been a lot of chaos there, and I wanted to highlight it. This is a pretty turbulent space, as is typical for areas with a flood of interest. Are we coming through the trough of disillusionment? Sure feels like it.
Savvy providers are monitoring the fray and helping customers with their preparations, working to keep them insulated from any volatility.