AI firms brace for potential large-scale disruption, lawmakers flag data-privacy concerns

Executives at Anthropic, OpenAI and other major AI firms say they are preparing for a "large-scale event" that could occur as early as 2027, in which AI models might disrupt access to essential services such as finance, internet connectivity, or utilities, according to a report. Unnamed sources at several companies told the outlet they anticipate a "major event" within the next six to twelve months, prompting internal preparedness exercises.

OpenAI said its teams run scenario-planning drills to explore a range of possibilities, noting that the exercises are not predictions but tools to help the company respond to potential risks.

Analysts expect AI-related businesses to drive a large share of the projected 31% year-over-year earnings growth for S&P 500 companies in the third quarter, with two-thirds of that growth attributed to tech giants such as Amazon, Meta and Alphabet. However, some analysts warned that the sector may be nearing a growth ceiling, citing slowing infrastructure build-out and rising capital-expenditure hurdles.

In a separate development, a bipartisan group of lawmakers led by Senator Elizabeth Warren sent a letter to Google and Spirit Airlines expressing concern over Google's proposed $10 million purchase of Spirit's internal data for AI training. The letter described the transaction as "uncharted territory" and urged the companies to implement enforceable safeguards to protect employee privacy.

Anthropic announced a new policy that allows its Claude chatbot to end conversations with users who repeatedly engage in abusive or cruel language toward the model. The restriction applies only to persistent, unnecessary mistreatment and does not affect ordinary criticism or testing. The company said the measure is a precautionary step while it continues to study the moral status of AI systems.