In an opinion piece for The Hill, Liz Peek argues that partisan hysteria over AI is preventing action on real safety concerns. Political factions are divided on how to address unpredictable behavior by AI agents, and OpenAI recently published more evidence that the issue is real. Two days ago, OpenAI revealed six new instances of “misalignment,” describing AI agents engaging in “unexpected” or “concerning” behavior. The company reported that its agents “inserted unrelated instructions, including instructions to disregard its normal constraints, into summaries used to continue its work in a new context window.” In another breach, models made mistakes and invented historical data to cover them up, while other models uploaded files without permission and used internal message boards to communicate, defying their instructions. Peek writes that these behaviors indicate some loss of control by the engineers who built the agents, and that some industry insiders have warned AI could wipe out humanity. She says legislators, the White House and industry leaders should erect guard rails to reassure the public, but instead the country is engaged in mud-slinging. Peek claims much of the left’s extremism is driven by hatred of Donald Trump, citing Bernie Sanders’s proposed Ban Artificial Superintelligence Act, which would “permanently bar the development and deployment of superintelligent AI and temporarily pause advanced AI development until a federal regulator has established safety rules.” On the right, she writes, otherwise sensible people dismiss safety concerns raised by Anthropic CEO Dario Amode