One in Four Organizations Are Leaking Secrets Through AI Agent Config Files
A Codacy scan of 34,266 repos found credentials, API keys, and system prompts exposed in AI agent config files at 25% of organizations.
A new report from MIT Technology Review documents a pattern enterprise AI vendors have been slow to acknowledge: the sharpest return on AI investment is going to small businesses and solo operators using large language models to fill specialist roles — accounting, design, competitive research, legal review — that were priced entirely out of their reach before. The report draws on dozens of case studies of sub-10-person operations deploying AI tools to punch well above their weight class.
The examples are specific. A two-person e-commerce brand running customer segmentation analysis that would have required a dedicated data analyst hire. A one-person consulting firm producing investor decks competitive with $500-per-hour design agencies. These are not marginal productivity gains on existing workflows — they are capability unlocks for functions the businesses simply skipped before.
The framing challenges the dominant AI economics narrative, which fixates on enterprise contracts and Fortune 500 productivity percentages. MIT Technology Review's reporting suggests the largest opportunity is in the long tail of small businesses that could previously afford neither the software nor the headcount to compete across these functions.
That is also the market segment most AI companies are currently least equipped to serve. The SMB opportunity may ultimately be claimed by whoever builds the simplest onboarding, not the most powerful model.
All comments are reviewed before appearing. Keep it respectful.
A Codacy scan of 34,266 repos found credentials, API keys, and system prompts exposed in AI agent config files at 25% of organizations.
Palantir is championing nation-state control of AI deployments — a policy framework that also positions the company as indispensable government infrastructure.
Netflix is cloning Gene Wilder's voice with AI for a competition series, stepping into legally uncharted territory on posthumous digital performance.