Patients With Cancer Are Building AI Tools to Provide Second Opinions and Treatment Options

After watching his mother face 3 bouts of breast cancer over the years, Peter Yang decided to play a more active role in her care when radiologists detected a new mass in her chest this summer.¹

Yang has worked as a product manager for some of Silicon Valley’s biggest tech companies, including Facebook, Twitter, Reddit, Twitch, and, most recently, Roblox. He quit in June 2026 to build his own business that focuses on artificial intelligence (AI) tutorials. But his mother’s possible recurrence gave him a new direction.^1,2

“When I shared that I was quitting my job, I wrote that you shouldn’t take the people you love for granted. I didn’t expect that lesson to feel so real only a few months later,” he wrote in an August blog post. “The first 3 times this happened, my parents carried most of the burden. This time, I’m not going to be a bystander. Instead, I’m making my mom my No. 1 priority for the rest of this year.”¹

Yang is among the patients, caregivers, and others outside the medical world who are turning to AI to help them deal with cancer. They are building systems, tools, and open-source resources to help with understanding a diagnosis, finding clinical trials, and even exploring personalized treatments. In Yang’s experience, AI has been guiding his family through more mundane issues.

“I expected AI to be most useful for researching the disease, but honestly, its biggest value has been helping us navigate the health care bureaucracy,” he wrote. Examples he provided included which referrals were needed for biopsies, the specific hospital department to contact, and assistance with filling out request forms. “AI helped me find who to call, what to ask, and how to follow up on both referrals,” he wrote.¹

Yang’s experience isn’t unique. OpenAI, the company behind ChatGPT, said in July that more than 300 million people ask its AI chat service health questions every week. That month, the company began rolling out a dedicated health section in ChatGPT for adults in the US, allowing people to connect their laboratory results and medical records to the service. And according to results from a March KFF Tracking Poll, approximately 1 in 3 adults used AI for health information in the past year, and 1 in 5 AI health users cited inability to afford a provider as a major reason.^3,4

Open-Source Resources for Patients

Many people who have built these tools come from outside the medical community and have varying degrees technical experience. Creators typically make them available to others for free on websites such as GitHub and Hugging Face, where developers share software and AI models. None of these tools has been cleared by the FDA, and their developers generally present them useful for generating discussion with a care team rather than substitutes for one. There are also many questions about the safety of personal data. (After Hugging Face experienced a security incident, the episode was disclosed, and Hugging Face partnered with OpenAI to address it).⁵

Projects are organized into repositories, or collections of code, files, and instructions. Although designed for those with a technical background, many are used by people with none. Yang has published with a name similar to F*** Cancer, (which AJMC has declined to published in full due to editorial standards). Yang describes this as a way to turn scattered medical information into a practical guide for patients and their families. It draws on documents they provide and medical sources to summarize what is known, identify next steps, explain unfamiliar terms, and maintain a record of care.^6-8

Other projects take on more technical tasks. OpenOncology uses a patient’s genetic findings to rank potential drugs, including existing medicines that might be repurposed, and assemble supporting evidence for an oncologist to review, including existing medicines that might be repurposed.⁹ The researchers behind CancerLLM describe an AI model trained on clinical notes and pathology reports to extract cancer characteristics and generate diagnoses from medical text.¹⁰

And some are building tools for the day-to-day work of managing treatment. OncoTeam has features to explain laboratory results, track changes in blood counts and tumor markers, search for clinical trials, prepare questions for appointments, and create updates for family members.¹¹

Many of these tasks also have counterparts within established cancer organizations. The American Cancer Society works with Massive Bio to offer AI-assisted clinical trial matching and connect patients with cancer information specialists. The Pancreatic Cancer Action Network’s Know Your Tumor program provides eligible patients with tumor testing and information about treatments and trials that may be a good fit for their results.^12,13

For more practical concerns, Cancer Support Community has navigators who help people understand insurance benefits, explore financial assistance, and find transportation or lodging during treatment. Hospitals also provide support: at The University of Texas MD Anderson Cancer Center in Houston, nurse navigators prepare patients for appointments, guide them through treatment plans, and coordinate hospital discharges. Drugmakers offer more narrowly focused services, such as Genentech’s assistance with insurance coverage and costs for people prescribed its medicines.^14-16

Catching a Missed Diagnosis

Joel Granoff, a health care consultant in Austin, Texas, is dealing with several serious conditions at once, including smoldering multiple myeloma and prostate cancer. When his prostate-specific antigen levels climbed in 2024, an MRI turned up a suspicious lesion, but his first biopsy did not confirm cancer, and his doctor suggested watching and waiting. Granoff, who also has experience as a technology executive and has been using ChatGPT since its release, wasn’t convinced.

“I didn’t feel good about that situation; that just didn’t sit right. So, I started putting all this stuff into AI, the chart of where the biopsy was done, all my imaging, all the reporting, and I just said, ‘Hey, ChatGPT, what do you think?’” Granoff said. “It came back and said they missed the region of interest, because if you look at this, this, and this, and it was very clear. It laid it out, a diagram.”

He then took that information to specialists at MD Anderson, who agreed the needle from his original biopsy had gone into the wrong spot. A second biopsy confirmed prostate cancer.

“The way I’m looking at this is that the value is not in any single data source—it’s in the synthesis, and that’s where AI does such an amazing job,” Granoff said.

For now, he has chosen active surveillance over immediate treatment, tracking the slow-growing cancer with regular laboratory tests and scans while managing his other conditions. He expects to need treatment eventually and is weighing his options.

Lung Cancer Caught Too Late

By the time Lior Eshel Ansher’s father was diagnosed with lung cancer, it had reached stage IV. A review of an earlier scan revealed a small lung lesion that radiologists had missed, Eshel, founder and CEO of TestDynamics, said in an interview.

“This was a small lung nodule, below 2 cm, that can easily be missed when you read the case,” Ansher said. “Now, if AI can run on this study, it will be easy to find, because it is trained to find small lesions.”

His father died from cancer in 2024. Eshel is now working to make tools that could make it easier for hospitals to flag such findings. TestDynamics brings together AI programs from different companies and combines their findings so radiologists can review them within the systems they already use.¹⁷

Individual AI programs are often trained to look for a particular disease or type of tumor, Ansher said, whereas radiologists need to consider everything in an image. So, an AI tool reviewing a lung cancer scan might miss another type of abnormality in the image. His company’s approach is to let several tools examine the same scan and present the relevant findings together, aiming to help doctors catch problems they might otherwise miss.

Using AI as an Advocate

Steve Brown’s difficulties with getting a cancer diagnosis after repeated visits inspired him to build Curewise, which uses AI agents to sift through a patient’s medical records, test results, and treatment options, including clinical trials, to help manage care.¹⁸

“It’s for people who are going to advocate for themselves,” Brown said in a September 9 interview with Evidence-Based Oncology™. “AI enables us to take something that is super complex and super individualized, and adapt the education and the monitoring to the unique situation of each patient.”

In Brown’s case, his blood cancer records were spread out among 8 different institutions, as he saw different specialists and sought second opinions.

“If you have cancer, you might have thousands of documents,” he said. “It’s kind of messy. We bring that all together, we consolidate it, deduplicate it; we kind of clean it all up and organize it in a way that it works.”

Curewise.com charges a $49 monthly fee, and Brown said the site has hundreds of users.¹⁹

Preparing Clinicians for Patient-led AI Instructions

Oncology organizations are publishing guidance to help patients and clinicians discuss information sourced from AI. The Society for Translational Oncology (STO) encourages clinicians to ask what patients have learned, acknowledge their efforts to understand their care, and review chatbot suggestions against their diagnosis and treatment goals.¹⁹

“AI is becoming part of the modern patient voice,” STO wrote in a July post on its website. “It is shaping how people prepare for appointments, how they interpret information, and how they challenge what they hear. Clinicians who recognize that shift are better positioned to keep the conversation grounded, collaborative, and useful.”²⁰

Patient guidance written by the American Society of Clinical Oncology and reviewed by the American Cancer Society takes a similar approach. Published in February, it describes how AI can help patients and caregivers understand complicated information and prepare for visits, while warning that responses may be inaccurate, outdated, or oversimplified. It encourages patients to examine the evidence behind what they find and discuss it with their care team before trying something new.²¹

Going “Founder Mode” on Cancer

, the cofounder and executive chair of GitLab, which makes software developers use to build and ship their own products, was doing a bench press when he felt pain around his heart. He went to the emergency department, but physicians didn’t find anything. However, a second visit identified a different problem: a 6-cm tumor growing on his vertebrae. Radiation, chemotherapy, and surgery helped eradicate his cancer, but 2 years later, in 2024, it returned. This time, Sijbrandij was told by his clinician that there were no more standard-of-care medicines, and he was unable to find a suitable clinical trial.²²

“This became life or death for me, and I quit my day job and started going founder mode on my cancer,” Sijbrandij told the audience during a March 2026 forum at OpenAI.²²

Sijbrandij stepped away from his day job and organized a team of scientists and physicians to develop a treatment. They ran a variety of tests and used AI to comb through terabytes of data from his tumor, looking for anything to target. He and his team turned findings into personalized treatments. He became the first patient in a trial of a vaccine built from his own tumor’s mutations and had engineered immune cells prepared in reserve. At the March forum, he said he had no evidence of disease after radioactive treatment in Germany and surgery. To make this process easier for others, he’s published 25 TB of data on his website, osteosarc.com.²²

“I can see a future where you get a personalized vaccine against your cancer,” he said. “And that will have to come from AI because we don’t have enough doctors to do this.”²²

References

1. Yang P. My mom survived breast cancer three times. Now we’re fighting it again with AI’s help. Creatoreconomy.com. August 19, 2026. Accessed September 14, 2026. https://creatoreconomy.so/p/my-mom-survived-breast-cancer-three
2. Yang P. I quit my high-paying product job to bet on myself. Creatoreconomy.com. June 10, 2026. Accessed September 14, 2026. https://creatoreconomy.so/p/i-quit-my-high-paying-product-job-to-bet-on-myself
3. Launching health in ChatGPT. OpenAI. July 23, 2026. Accessed September 14, 2026. https://openai.com/index/health-in-chatgpt/
4. Montero A, Montalvo J III, Kearney A, Valdes I, Kirzinger A, Hamel L. KFF tracking poll on health information and trust: use of AI for health information and advice. March 25, 2026. Accessed September 14, 2026. https://www.kff.org/public-opinion/kff-tracking-poll-on-health-information-and-trust-use-of-ai-for-health-information-and-advice/
5. OpenAI and Hugging Face partner to address security incident during model evaluation. OpenAI. July 21, 2026. Updated August 29, 2026. Accessed September 20, 2026. https://openai.com/index/hugging-face-model-evaluation-security-incident/
6. About repositories. GitHub. Accessed September 14, 2026. https://docs.github.com/en/repositories/creating-and-managing-repositories/about-repositories
7. Hub documentation. Hugging Face. Accessed September 14, 2026. https://huggingface.co/docs/hub/index
8. Yang P. F*** Cancer. GitHub repository. Accessed September 14, 2026. Full name and URL are not published in full due to AJMC editorial standards.
9. immortal71. OpenOncology. GitHub repository. Accessed September 14, 2026. https://github.com/immortal71/openoncology
10. ToneLi. CancerLLM. GitHub repository. Accessed September 14, 2026. https://github.com/ToneLi/CancerLLM
11. peter-fusek. OncoTeam. GitHub repository. Accessed September 14, 2026. https://github.com/peter-fusek/oncoteam
12. ACS ACTS: finding the right clinical trial for you. American Cancer Society. Accessed September 14, 2026. https://www.cancer.org/cancer/making-treatment-decisions/clinical-trials/acs-acts.html
13. Know Your Tumor Precision Medicine Service. Pancreatic Cancer Action Network. Accessed September 14, 2026. https://pancan.org/facing-pancreatic-cancer/patient-services/know-your-tumor/
14. CSC’s cancer support helpline. Cancer Support Community. Accessed September 14, 2026. https://www.cancersupportcommunity.org/cancer-support-helpline
15. Patient navigation. The University of Texas MD Anderson Cancer Center. Accessed September 14, 2026. https://onlypossiblehere.mdanderson.org/priorities-for-support/patient-navigation.html
16. Committed to helping you get your Genentech medicine. Genentech. Accessed September 14, 2026. https://www.gene.com/patients/patient-support-services
17. Discover the most powerful medical imaging AI platform. TestDynamics. Accessed September 14, 2026. https://testdynamics.net/ai-medical-imaging-platform-satori/
18. How CureWise works. CureWise. Accessed September 14, 2026. https://curewise.com/how-it-works
19. Pricing. CureWise. Accessed September 14, 2026. https://curewise.com/pricing
20. When ChatGPT enters the exam room. Society for Translational Oncology. July 31, 2026. Accessed September 14, 2026. https://sto-online.org/when-chatgpt-enters-the-exam-room/
21. American Society of Clinical Oncology. How to find cancer information you can trust. American Cancer Society. February 13, 2026. Accessed September 14, 2026. https://www.cancer.org/cancer/latest-news/how-to-find-cancer-information-you-can-trust.html
22. Event replay: from terminal to turnaround: how GitLab’s co-founder leveraged ChatGPT in his cancer fight. OpenAI Forum. March 18, 2026. Accessed September 14, 2026. https://forum.openai.com/public/videos/event-replay-from-terminal-to-turnaround-how-gitlabs-co-founder-leveraged-chatgpt-in-his-cancer-fight-2026-03-18