AI news · Monday, June 22, 2026
OpenAI uses frontier models to patch critical open-source software vulnerabilities
OpenAI is launching an initiative called Patch the Planet, partnering with security firm Trail of Bits to use its AI models to find and fix bugs in foundational open-source software. The program focuses on critical infrastructure like the Linux kernel, Python, and web browsers, using AI-assisted research to identify and patch security holes. Trail of Bits engineers manually review every AI-generated finding before it hits maintainers to avoid overwhelming them with false positives. In initial tests, the team used models like GPT-5.5-Cyber to create entire fuzzing labs—systems that test for software errors—in less than a day, a task that typically takes humans weeks. The project has already identified hundreds of security issues, including a 23-year-old vulnerability in OpenBSD, and successfully blocked several Firefox exploits before they could be used in a major hacking competition.
Meanwhile, internal morale at Meta is hitting a new low following a massive security blunder. The company was forced to pause its controversial Model Capability Initiative—a program that tracked employee mouse clicks, keystrokes, and screen content to train AI—after a database containing sensitive worker data was exposed to the entire company. Employees had been petitioning against the invasive surveillance for months, and the leak validated concerns about the lack of privacy safeguards in the initiative. This incident follows a broader, often messy, corporate reorganization that saw 6,500 staff members moved into new AI-focused roles. While Meta continues to push for AI dominance, its hardware side is seeing a bizarre trend: a third-party startup called Lorika is selling $35–$40 clip-on polycarbonate frames to make Meta’s Ray-Ban smart glasses look "puffy" and more colorful, targeting users who find the standard designs too boring.
Finally, the AI-driven data center boom is creating a weird ethical divide among the skilled tradespeople responsible for building them. While tech giants are pouring billions into massive data centers and funding trade academies to train the next generation of workers, some electricians are pushing back. On forums like Reddit, workers debate whether building these facilities makes them complicit in corporate greed or the environmental cost of AI. Some refuse the work entirely, even at a pay cut, while others view it as a necessary way to secure a future in an economy they feel is increasingly dominated by a handful of AI companies. For those on the front lines of the acquisition market, the scrutiny is even more intense. Consulting firm Bain & Company is now using "vibecoding"—creating AI-generated replicas of software products—to determine if a company’s tech is actually defensible or just hype. The practice is already influencing massive financial decisions, contributing to a 69 percent collapse in private equity tech deals in early 2026 as investors grow terrified of the speed at which AI can render current software obsolete.
The quick hits
- OpenAI launched an initiative to use its AI models to identify and patch security vulnerabilities in critical open-source software like Linux and Python. — It helps automate the tedious work of finding and fixing software bugs, theoretically making the internet safer for everyone.
- Meta paused its employee-tracking program after a security lapse exposed sensitive keystroke and screen data to workers across the company. — It was an invasive monitoring project that employees had already been protesting, and the accidental leak of that data turned the internal situation into a full-blown morale crisis.
- Bain & Company is using AI to create software replicas to test if an acquisition target’s tech is actually valuable or easy to copy. — It is a new "vibecoding" reality check that is causing investors to pull out of deals as they realize how quickly AI can devalue existing software.
- Electricians are openly debating whether building AI data centers makes them "sellouts" to corporate interests. — It shows how the massive physical construction required for the AI boom is creating a real-world labor backlash among the people actually doing the work.