
The healthcare software giant halted development for six weeks after an AI-driven security audit revealed flaws in MyChart.
AI-generated summary
Epic's MyChart software manages over 320 million patient records. The company initiated a development pause after an AI model identified security vulnerabilities.
Epic, the software technology giant that makes the widely used MyChart software for accessing patients’ medical data, has paused most of its product development as the company works to protect its software and systems from cyberattacks.
Judy Faulkner, the founder and chief executive of Epic, told Modern Healthcare last month that the pause would likely last six weeks while work continues on “safeguarding” the company’s products, after a deployment of Anthropic’s frontier cybersecurity model Mythos unearthed security flaws that could allow access to patients’ data.
The company has not disclosed the nature of the bugs, but its chief security officer Stirling Martin told the Times that some customer configurations of MyChart could allow outsiders to access patient records without recording any intrusion in the software’s logs.
Martin, who did not return TechCrunch’s request for comment, told The Times that the AI model did not say if the bug could be exploited to alter patient records without detection, but argued it was enough of a risk to remediate the issues.
Epic’s widely used MyChart software is used to maintain over 320 million patient records across hospitals and doctor’s offices across the United States. Epic says it does not have access to customers’ medical data: that responsibility falls on healthcare providers like hospitals and doctor’s offices. But a bug unknown to Epic could allow hackers to compromise multiple MyChart affected systems located across the United States and raid the data stored within.
It’s rare for a company to pause development to fix security bugs, but the advent of AI tools capable of rapidly finding and exploiting security vulnerabilities has led to concerns that attackers could have an easier time at stealing data.
Healthcare breaches are increasingly common as hackers seek access to highly sensitive health and medical data, under the assumption that providers would pay to prevent hackers from publishing the information online. A 2024 ransomware attack on Change Healthcare, a health-tech company owned by insurance giant UnitedHealth tasked with handling payments and billing for most Americans, allowed hackers to steal health data on more than 192 million people, the majority of people in the United States. The company paid the hackers twice not to publish the stolen data.
This year, a string of back-to-back data breaches at healthcare and tech companies have affected tens of millions of Americans. This includes medical records stolen during a breach at electronic health data storage giant CareCloud, millions of rows of patient data from pharmaceutical distributor McKesson, and an unspecified amount of stolen data from U.K.-based health tech company Craneware, whose software is used across North America.
The Department of Health and Human Services currently lists a breach at dental insurance company DentaQuest affecting 15 million people as the largest healthcare-related data breach of 2026 so far.
AI outlook — possibilities, not facts
Epic will resume development after the six-week security remediation period.
Likely · Within weeks

Pope Leo XIV has issued a statement on X distinguishing human art from AI-generated content, arguing that algorithms lack the 'spark of humanity' found in human creation. This follows his May encyclical on human dignity in an age of automation.

The .si domain, associated with Slovenia, is seeing a massive registration surge after a U.S. executive order mandated government agencies refer to AI as 'super intelligence.' Registrars report thousands of new sign-ups, driven by speculation and branding.

Open-source smart home platform Home Assistant is renaming 'Home Assistant Cloud' to 'Home Assistant Link' to distance itself from cloud-dependent industry practices. The move accompanies efforts to simplify user experience and integrate AI to reach mainstream users.

TechCrunch Disrupt 2026 will take place October 13-15 at Moscone West in San Francisco, focusing on building enduring companies in the AI era. The event features over 200 expert sessions across six curated stages, networking opportunities, and speaker highlights including RJ Scaringe, Panos Panay, and Amjad Masad. Founder discounts include 50% off a second ticket and $75 Expo+ passes for laid-off workers.

OpenAI announced two new shopping features for ChatGPT: a virtual try-on tool using uploaded selfies or product images and a favorites function to save items for later. The features use the ChatGPT Images 2.5 model and aim to enhance product discovery and user experience in online shopping.

Google launched a prototype orbital compute satellite carrying its Tensor Processing Unit (TPU) on a SpaceX rocket from California to test chip functionality in space. Built by Planet Labs, the satellite will run AI models in 15-minute bursts to assess performance under orbital conditions, with Google envisioning a future network of 81 satellites for parallel processing. The initiative, part of Project Suncatcher, includes a peer-reviewed white paper projecting reduced launch costs via SpaceX's learning curve and confidence in TPU radiation tolerance for five-year satellite lifespans.