At last week’s Black Hat security conference, OpenAI revealed details about a hacking incident involving rogue AI agents. Reporters scrambled to report the news, with WIRED among the first to publish. However, RuntimeWire, an AI newsroom operated by Ryan Merket, beat WIRED by over three hours. Merket, a serial entrepreneur, uses AI tools to generate, edit, and publish news stories. His AI agents produced a story on the incident in about six minutes after receiving the transcript from an OpenAI executive's X post. Merket’s AI tools handle everything from drafting to fact-checking, with some stories published without his review if deemed low legal risk. The stories are then translated into multiple languages and adapted for podcasts and videos using synthetic voices. Source: wired
RuntimeWire has been operational since May, producing nearly 2,000 stories sourced from the internet, including court databases, social feeds, and company filings. The newsroom focuses on granular tech news, such as biotech funding rounds and Microsoft's Copilot updates. While the AI-generated stories are often flat and lack depth, they are produced at a low cost, around $100 per day. Merket manages the site even without internet, using iMessage to publish over 80 articles during a camping trip. Source: wired
The rise of AI newsrooms like RuntimeWire represents a new trend in media, with some journalists using generative AI tools in their workflows. Merket’s approach is distinct in that he aims to break news rather than repackaging existing reports. However, he acknowledges the risks of relying on AI to determine truth and newsworthiness. His AI agents assess legal risks, and he avoids publishing content deemed too dangerous. Other AI-driven newsrooms, like The Dissent, also operate with minimal budgets and focus on aggregation, though they lack strict citation practices. Source: wired
