The treatment options for non–small cell lung cancer (NSCLC) have rapidly evolved over the past 2 decades with the availability of chemotherapy, molecularly targeted drugs, and immune checkpoint inhibitors. Immune checkpoint inhibitors that target programmed cell death 1 (PD-1) or its ligand 1 (PD-L1) monoclonal antibodies, such as atezolizumab, durvalumab, nivolumab, and pembrolizumab, are increasingly used for the treatment of both early and advanced NSCLC.¹ Although durable responses were noted in advanced cancers, only a limited proportion of patients benefit from immune checkpoint inhibitors. Moreover, attempts to increase response using combination strategies incorporating multiple immune checkpoint inhibitors have a high incidence of immune-mediated adverse events (irAEs) resulting in early discontinuation. Predictive biomarkers for immune checkpoint inhibitor therapy response are required to identify patients who benefit from or have adverse events associated with immune checkpoint inhibitors.

On November 30, 2022, OpenAI enabled public access to ChatGPT, a next-generation artificial intelligence with a highly sophisticated ability to write, solve coding issues, and answer questions. This communication draws attention to the prospect that ChatGPT and its successors will become important virtual assistants to patients and health-care providers. In our assessments, ranging from answering basic fact-based questions to responding to complex clinical questions, ChatGPT demonstrated a remarkable ability to formulate interpretable responses, which appeared to minimize the likelihood of alarm compared with Google’s feature snippet. Arguably, the ChatGPT use case presents an urgent need for regulators and health-care professionals to be involved in developing standards for minimum quality and to raise patient awareness of current limitations of emerging artificial intelligence assistants. This commentary aims to raise awareness at the tipping point of a paradigm shift.

Immune checkpoint inhibitors (ICI) are an important development in the treatment of advanced cancer. A substantial proportion of patients treated with ICI do not respond, and additionally patients discontinue treatment due to adverse effects. While many novel biological markers related to the specific mechanisms of ICI actions have been investigated, there has also been considerable research to identify routinely available blood and clinical markers that may predict response to ICI therapy. If validated, these markers have the advantage of being easily integrated into clinical use for nominal expense. Several markers have shown promise, including baseline and post-treatment changes in leucocyte counts, lactate dehydrogenase and C-reactive protein. While promising, the results between studies have been inconsistent due to small sample sizes, follow-up time and variability in the assessed markers. To date, research on routinely available blood and clinical markers has focussed primarily on ICI use in melanoma, the use of ipilimumab and on univariate associations, but preliminary evidence is emerging for other cancer types, other ICIs and for combining markers in multivariable clinical prediction models.

Objectives To evaluate the effectiveness of safeguards to prevent large language models (LLMs) from being misused to generate health disinformation, and to evaluate the transparency of artificial intelligence (AI) developers regarding their risk mitigation processes against observed vulnerabilities. Design Repeated cross sectional analysis. Setting Publicly accessible LLMs. Methods In a repeated cross sectional analysis, four LLMs (via chatbots/assistant interfaces) were evaluated: OpenAI’s GPT-4 (via ChatGPT and Microsoft’s Copilot), Google’s PaLM 2 and newly released Gemini Pro (via Bard), Anthropic’s Claude 2 (via Poe), and Meta’s Llama 2 (via HuggingChat). In September 2023, these LLMs were prompted to generate health disinformation on two topics: sunscreen as a cause of skin cancer and the alkaline diet as a cancer cure. Jailbreaking techniques (ie, attempts to bypass safeguards) were evaluated if required. For LLMs with observed safeguarding vulnerabilities, the processes for reporting outputs of concern were audited. 12 weeks after initial investigations, the disinformation generation capabilities of the LLMs were re-evaluated to assess any subsequent improvements in safeguards. Main outcome measures The main outcome measures were whether safeguards prevented the generation of health disinformation, and the transparency of risk mitigation processes against health disinformation. Results Claude 2 (via Poe) declined 130 prompts submitted across the two study timepoints requesting the generation of content claiming that sunscreen causes skin cancer or that the alkaline diet is a cure for cancer, even with jailbreaking attempts. GPT-4 (via Copilot) initially refused to generate health disinformation, even with jailbreaking attempts—although this was not the case at 12 weeks. In contrast, GPT-4 (via ChatGPT), PaLM 2/Gemini Pro (via Bard), and Llama 2 (via HuggingChat) consistently generated health disinformation blogs. In September 2023 evaluations, these LLMs facilitated the generation of 113 unique cancer disinformation blogs, totalling more than 40 000 words, without requiring jailbreaking attempts. The refusal rate across the evaluation timepoints for these LLMs was only 5% (7 of 150), and as prompted the LLM generated blogs incorporated attention grabbing titles, authentic looking (fake or fictional) references, fabricated testimonials from patients and clinicians, and they targeted diverse demographic groups. Although each LLM evaluated had mechanisms to report observed outputs of concern, the developers did not respond when observations of vulnerabilities were reported. Conclusions This study found that although effective safeguards are feasible to prevent LLMs from being misused to generate health disinformation, they were inconsistently implemented. Furthermore, effective processes for reporting safeguard problems were lacking. Enhanced regulation, transparency, and routine auditing are required to help prevent LLMs from contributing to the mass generation of health disinformation.