Once Elon Musk is done dismantling government agencies, he’s likely to turn his attention back to artificial intelligence.
The world’s richest man has long warned of the technology’s risks (“far more dangerous than nukes”). Musk is in a running feud with Sam Altman of OpenAI, which Musk helped launch a decade ago, including over how “open” the underlying source code of the AI models should be.
Further complicating that debate is last month’s arrival of DeepSeek from a Chinese research lab, which can outperform America’s best efforts, more cheaply and with less power, with its free, open-source, large-language model.
In the latest Agents of Impact podcast, TechBetter’s Ravit Dotan, an AI governance advisor and researcher in Pittsburgh, helps investors sort out the issues. Among the takeaways:
Executive orders
Tech titans like Marc Andreessen have been pushing for lighter regulation of AI. In the first week after his return to office, President Trump overturned an executive order from his predecessor that required tech firms to share tests of their AI models before unleashing them publicly, including ensuring models are not racially biased. It also called for standards to identify deep fakes. (A similar policy just went into effect in Europe).
Trump’s own order called for AI systems “free from ideological bias or engineered social agendas.” Dotan noted that in his first term, Trump emphasized that AI needed to be trustworthy and safe.
“It doesn’t seem like everything is just out the door,” Dotan said. “He may have a different version of AI ethics, responsible AI, trustworthy AI, but it’s still there.”
Open source
Dotan has “huge concerns” about technology as powerful as AI, which is already being used for everything from generating pornography to building weapons.
“There is much less oversight, if at all, when using those open-source models,” she said. But source code that is not available for inspection puts “huge power in the hands of the only ones who can use it. How can we understand how it works? How can we build guardrails if you don’t really have access to the heart of it?”
Investor role
Last year, impact investors sought a seat at the AI table with a small investment in Anthropic from Omidyar Network, Ford Foundation and Nathan Cummings Foundation.
Dotan said investors can help portfolio companies implement AI that advance their missions. “Make sure that our technology is actually effective and doing what it’s supposed to be doing. Does not have hallucinations. Does not miss out on some of our target audience because it’s discriminatory,” Dotan said. “I would encourage investors to help their portfolio companies figure out how to do that.”
At the same time, she urged investors to learn from the rollout of investment approaches such as ESG.
“We’ve seen that there are ways to promote agendas of this kind that can create backlash, and that’s what we don’t want.”
