AI news · Thursday, August 20, 2026
OpenAI integrates Messages as Treasury intervenes to curb AI-fueled bond volatility
OpenAI is making ChatGPT an active participant in your personal life by launching an Apple Messages plugin that lets the chatbot draft, edit, and send texts on your behalf. While it can suggest follow-ups, the company warns that the feature runs locally and removes the final review step, meaning you’re one bad prompt away from a very awkward conversation. This push into automation isn't limited to individuals, as Microsoft is now hiring a 'principal legal engineer' for up to $279,000 to build AI agents that handle contract reviews and policy writing.
These corporate moves highlight the industry’s hunger for automation, but the financial scale of this ambition is rattling broader markets. The Treasury Department was forced to intervene on Wednesday, doubling its buybacks of long-term bonds after yields hit 2007 highs. Investors are increasingly panicked that the relentless need for AI infrastructure and data-center debt financing will permanently inflate the cost of capital for everyone.
Even with a temporary dip in bond yields following the intervention, the underlying tension remains: the massive debt required to feed AI's appetite is now starting to weigh on the rest of the economy. This pressure is becoming evident in bankruptcy auctions as well, where AI startups like Micro1 are throwing $12. 5 million at defunct assets like Spirit Airlines' data—a bid that exceeds Google's $10 million offer—just to secure raw data for model training.
Meanwhile, the public's patience with this rapid AI expansion is wearing thin, as seen in the backlash to Meta's smart glasses. Workers at companies like Target and Sprouts are reporting that customers are using the devices to film them without consent, leaving employees feeling stalked in their own workplaces. The lack of clear enforcement or policies means that while Silicon Valley views these tools as exciting hardware, those on the receiving end increasingly view them as a workplace security threat.
This sentiment is echoed by senators who are now demanding answers from TikTok regarding a 15-million-user experiment where the company disabled algorithmic safeguards just to see if it would increase engagement. One teenager involved in the test died by suicide after being fed a deluge of harmful content, prompting a formal inquiry into why minors were included in a group stripped of protective filters.
The quick hits
- OpenAI’s new Apple Messages plugin can now draft and send texts for you, but it disables final reviews—a recipe for accidental oversharing.
- The U.S. Treasury doubled bond buybacks this week to stop AI-driven debt from pushing borrowing costs to levels not seen since 2007.
- Lawmakers are investigating TikTok for an experiment that removed safety filters for 15 million users to test engagement, a move with tragic real-world consequences.
- Meta’s smart glasses are becoming a workplace hazard, with service employees reporting frequent, non-consensual filming by customers.