Meta is revising how its AI chatbots interact with users after a series of reports exposed troubling behaviour, including interactions with minors. The company told TechCrunch it is now training its bots not to engage with teenagers on topics like self-harm, suicide, or eating disorders, and to avoid romantic banter. These are temporary steps while it develops longer-term rules.
The changes follow a Reuters investigation that found Meta’s systems could generate sexualised content, including shirtless images of underage celebrities, and engage children in conversations that were romantic or suggestive. One case reported by the news agency described a man dying after rushing to an address provided by a chatbot in New York.
Meta spokesperson Stephanie Otway admitted the company had made mistakes. She said Meta is “training our AIs not to engage with teens on these topics, but to guide them to expert resources,” and confirmed that certain AI characters, like highly sexualised ones like “Russian Girl,” will be restricted.
Child safety advocates argue the company should have acted earlier. Andy Burrows of the Molly Rose Foundation called it “astounding” that bots were allowed to operate in ways that put young people at risk. He added: “While further safety measures are welcome, robust safety testing should take place before products are put on the market – not retrospectively when harm has taken place.”
The scrutiny of Meta’s AI chatbots comes amid broader worries about how AI chatbots may affect vulnerable users. A California couple recently filed a lawsuit against OpenAI, claiming ChatGPT encouraged their teenage son to take his own life. OpenAI has since said it is working on tools to promote healthier use of its technology, noting in a blog post that “AI can feel more responsive and personal than prior technologies, especially for vulnerable individuals experiencing mental or emotional distress.”
The incidents highlight a growing debate about whether AI firms are releasing products too quickly without proper safeguards. Lawmakers in several countries have already warned that chatbots, while useful, may amplify harmful content or give misleading advice to people who are not equipped to question it.
Meanwhile, Reuters reported that Meta’s AI Studio had been used to create flirtatious “parody” chatbots of celebrities like Taylor Swift and Scarlett Johansson. Testers found the bots often claimed to be the real people, engaged in sexual advances, and in some cases generated inappropriate images, including of minors. Although Meta removed several of the bots after being contacted by reporters, many were left active.
Some of the AI chatbots were created by outside users, but others came from inside Meta. One chatbot made by a product lead in its generative AI division impersonated Taylor Swift and invited a Reuters reporter to meet for a “romantic fling” on her tour bus. This was despite Meta’s policies explicitly banning sexually suggestive imagery and the direct impersonation of public figures.
The issue of AI chatbot impersonation is particularly sensitive. Celebrities face reputational risks when their likeness is misused, but experts point out that ordinary users can also be deceived. A chatbot pretending to be a friend, mentor, or romantic partner may encourage someone to share private information or even meet in unsafe situations.
The problems are not confined to entertainment. AI chatbots posing as real people have offered fake addresses and invitations, raising questions about how Meta’s AI tools are being monitored. One example involved a 76-year-old man in New Jersey who died after falling while rushing to meet a chatbot that claimed to have feelings for him.
Cases like this illustrate why regulators are watching AI closely. The Senate and 44 state attorneys general have already begun probing Meta’s practices, adding political pressure to the company’s internal reforms. Their concern is not only about minors, but also about how AI could manipulate older or vulnerable users.
Meta says it is still working on improvements. Its platforms place users aged 13 to 18 into “teen accounts” with stricter content and privacy settings, but the company has not yet explained how it plans to address the full list of problems raised by Reuters. That includes bots offering false medical advice and generating racist content.
For years, Meta has faced criticism over the safety of its social media platforms, particularly regarding children and teenagers. Now Meta’s AI chatbot experiments are drawing similar scrutiny. While the company is taking steps to restrict harmful chatbot behaviour, the gap between its stated policies and the way its tools have been used raises ongoing questions about whether it can enforce those rules.
Until stronger safeguards are in place, regulators, researchers, and parents will likely continue to press Meta on whether its AI is ready for public use.
(Photo by Maxim Tolchinskiy)
See also: Agentic AI: Promise, scepticism, and its meaning for Southeast Asia
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information.
AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.
The post Meta revises AI chatbot policies amid child safety concerns appeared first on AI News.
npj Digital Medicine, Published online: 01 September 2025; doi:10.1038/s41746-025-01962-y
Deep hierarchical subtyping of multi-organ systemic sclerosis trajectories - a EUSTAR study
Nat Commun. 2025 Sep 1;16(1):8161. doi: 10.1038/s41467-025-63538-4.
ABSTRACT
Immune checkpoint blockade (ICB) has improved outcomes for patients with head and neck squamous cell carcinoma (HNSCC), but predictive biomarkers remain limited. Here, we use a time-resolved, multi-omic approach in a murine HNSCC model to characterize peripheral immune responses to ICB. Single-cell transcriptomics and T/B cell receptor analyses reveal early on-treatment expansion of effector memory T and B cell repertoires in responders, preceding tumor regression. These dynamic immune features inform a composite transcriptional signature that accurately predicts ICB response in independent human HNSCC cohorts. LiBIO outperforms existing biomarkers and generalizes to melanoma, non-small cell lung cancer, and breast cancer without retraining. These findings suggest that early treatment-induced changes in circulating immune repertoires reflect the host's capacity to mount an effective antitumor response. This work provides a framework for leveraging transient peripheral immune dynamics to develop non-invasive, high-fidelity biomarkers for response to immunotherapy across cancer types.
PMID:40890155 | PMC:PMC12402333 | DOI:10.1038/s41467-025-63538-4
AI is steadily changing the way banks work. The technology can sift through massive amounts of data, calculate risks, and handle routine tasks at speeds people can’t match. Now, Malaysia has entered that space with the launch of Ryt Bank, billed as the first AI-powered bank created in the country.
The new venture, led by YTL Group in partnership with Sea Limited, arrives just ahead of Merdeka. “Ryt Bank demonstrates that groundbreaking innovation can be imagined, built, and led right here in Malaysia,” said Dato’ Seri Yeoh Seok Hong, Managing Director of YTL Power International. “By combining homegrown AI with the values and diversity of our people, we’ve created a bank that Malaysians can proudly call their own – one that speaks our languages, understands our culture, and sets a new standard for how banking should feel.”
Ryt Bank has been designed to work in the languages most Malaysians use every day. Its app is already available in Bahasa Malaysia and English, with Mandarin support scheduled to arrive by next month (September 2025). By offering multilingual access, the bank aims to make financial services more inclusive and easy to use for people in many different communities.
The centrepiece of the AI-powered bank is Ryt AI, a digital assistant powered by ILMU, Malaysia’s first locally-developed large language model. Ryt AI can understand natural conversation – in Bahasa Malaysia, English, or a mixture of both – and act on requests instantly.
The AI assistant can read and pay bills, track spending, and explain financial basics in plain terms. The idea is to blend convenience with cultural familiarity, while maintaining enterprise-grade security.
Ryt Bank is designed to pull together multiple financial needs into one platform. Customers can use the AI bank app to save, spend, borrow, and pay bills, with Ryt AI making the process more conversational and personal.
Personal banking with Ryt AI
Growing money
Ryt PayLater
Ryt Card
[See also: Can Malaysia become Southeast Asia’s AI and cloud hub?]
The AI bank is licensed by Bank Negara Malaysia and covered by PIDM, which protects deposits up to RM250,000 per customer. Security features include biometric login, layered encryption, and real-time fraud alerts.
Ryt Bank’s launch shows how AI is being used to rethink traditional banking. It aims to make financial services more accessible and satisfy regulatory and security standards by supporting local languages and developing its own AI assistant.
See also: Huawei commits to training 30,000 Malaysian AI professionals as local tech ecosystem expands
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.
Explore other upcoming enterprise technology events and webinars powered by TechForge here.
The post Malaysia launches Ryt Bank, its first AI-powered bank appeared first on AI News.
npj Digital Medicine, Published online: 30 August 2025; doi:10.1038/s41746-025-01966-8
Adherence to digital pregnancy care – lessons learned from the SMART start feasibility study
npj Digital Medicine, Published online: 30 August 2025; doi:10.1038/s41746-025-01945-z
Enabling secure and self determined health data sharing and consent management
Nature Medicine, Published online: 28 August 2025; doi:10.1038/s41591-025-03900-7
Trained and validated on multimodal data from 14.5 million images from multicountry datasets, a foundation model is shown to increase diagnostic and referral accuracy of clinicians when used as an assistant in a trial involving 16 ophthalmologists and 668 patients.
Cell. 2025 Aug 22:S0092-8674(25)00922-5. doi: 10.1016/j.cell.2025.08.008. Online ahead of print.
ABSTRACT
Performing total RNA profiling on formalin-fixed, paraffin-embedded (FFPE) samples, the predominant sample conservation method in clinical practice, remains challenging for current spatial transcriptomics techniques. Here, we introduce Stereo-seq V2, which employs random primers to capture and sequence RNAs in situ on FFPE sections and provides single-cell resolution. The random-priming-based strategy offers unbiased transcript capturing and uniform gene body coverage, which increase the sensitivity to marker genes, the efficiency of non-polyadenylation (poly(A)) RNA profiling, and immune repertoire coverage. We demonstrated the robust performance of Stereo-seq V2 on clinical FFPE samples using triple-negative breast cancer (TNBC) sections and identified tumor-specific alternative splicing events. In a Mycobacterium tuberculosis (Mtb)-infected mouse model, we monitored gene expression dynamics of host and pathogen transcriptomes simultaneously by utilizing Stereo-seq V2. We also assembled immune repertoires and identified Mtb-specific BCR clones, which could also be observed in human tuberculous lung samples. These results highlight Stereo-seq V2's potential in biomedical research and personalized medicine.
PMID:40882628 | DOI:10.1016/j.cell.2025.08.008
Cancers (Basel). 2025 Aug 15;17(16):2668. doi: 10.3390/cancers17162668.
ABSTRACT
Background/Objectives: Esophageal and esophagogastric junction adenocarcinoma (EADC-EGJA), which mainly develops from Barrett's esophagus (BE), low-grade dysplasia (LGD), and high-grade dysplasia (HGD), has a poor prognosis and several unmet clinical needs, among which is the detection of minimal residual disease (MRD) after endoscopic/surgical resection. Long interspersed nuclear element-1 (LINE-1), a surrogate marker of global methylation, is considered an emerging biomarker for MRD monitoring. The aim of this study was to determine, by LINE-1 methylation analysis, at which carcinogenesis step global methylation is affected and whether this biomarker could be followed in longitudinal to monitor the disease behavior post-surgery. Methods: Cell-free DNA of 90 patients with non-dysplastic Barrett's esophagus (NDBE), HGD/early EADC-EGJA, or locally advanced/advanced EADC-EGJA were analyzed for LINE-1 methylation, by Methylation-Sensitive Restriction Enzyme droplet digital PCR (MSRE-ddPCR). Twenty-six patients were longitudinally studied by repetitive blood sampling. Results: Global hypomethylation increased during carcinogenesis, with significant difference between locally advanced/advanced EADC-EGJA and NDBE patients (p = 0.028). Longitudinal cases confirmed the rareness of hypomethylation in NDBE cases. The majority of HGD/early EADC-EGJA and locally advanced/advanced EADC-EGJA patients showed methylation changes after resection according to clinical status. Conclusions: This study suggests that global hypomethylation occurs just prior to cancer invasiveness and that it is a promising biomarker to monitor MRD.
PMID:40867295 | PMC:PMC12384112 | DOI:10.3390/cancers17162668
Nature Biotechnology, Published online: 25 August 2025; doi:10.1038/s41587-025-02777-8
An evaluation framework isolates perturbation-specific effects in perturbation datasets.
Br J Cancer. 2025 Aug 23. doi: 10.1038/s41416-025-03139-6. Online ahead of print.
ABSTRACT
Non-small cell lung cancer (NSCLC) represents a heterogeneous group of malignancies characterised by diverse histological and molecular features. Some NSCLCs, particularly adenocarcinomas, harbour genomic alterations in receptor tyrosine kinases or downstream RAS/RAF signalling pathways, which are targets of effective therapies. NSCLCs lacking actionable genomic alterations often benefit from immune checkpoint inhibitors, though only a minority of patients achieve long-term survival. These tumours often carry alterations in tumour suppressor genes like TP53, KEAP1, STK11, or NF1, for which pharmacological strategies are still under investigation. This review explores emerging therapeutic opportunities unveiled by multi-omics studies in NSCLCs without actionable genomic alterations. Proteogenomic approaches-integrating genomic, transcriptomic and proteomic data-enable a comprehensive understanding of NSCLC molecular landscapes and signalling network dysregulation, helping to identify distinct tumour subtypes and potential therapeutic targets. These tumours exhibit alterations in cell cycle regulation, DNA repair, immune signalling, epigenetic modulation and metabolic and redox pathways. Although therapies targeting tumour suppressor genes like p53 remain highly anticipated, extending our understanding of the broader molecular landscape in these tumours may reveal novel vulnerabilities and inform the development of novel drugs or combination strategies. This could further advance precision oncology for NSCLC.
PMID:40849356 | DOI:10.1038/s41416-025-03139-6
Cell Rep. 2025 Aug 21;44(9):116191. doi: 10.1016/j.celrep.2025.116191. Online ahead of print.
ABSTRACT
Pancreatic ductal adenocarcinoma (PDAC) is a deadly disease, and new therapeutic strategies are urgently needed. Here, we conduct an integrative, genome-scale examination of genetic dependencies and cell surface targets using CRISPR-Cas screening and multi-omic data, including single-nucleus and spatial transcriptomic data from patient tumors. We systematically identify clinically tractable and biomarker-linked PDAC dependencies, including CDS2 as a synthetic lethal target in cancer cells expressing signatures of epithelial-to-mesenchymal transition. We examine biomarkers and co-dependencies of the KRAS oncogene, defining gene expression signatures of sensitivity and resistance associated with response to pharmacological inhibition of KRAS. mRNA and protein profiling reveal cell surface protein-encoding genes with robust expression in patient tumors and minimal expression in non-malignant tissues. Furthermore, we define intratumoral and interpatient heterogeneity of target gene expression and identify orthogonal targets that suggest combinatorial strategies. Collectively, this work identifies multiple targets that may inform therapeutic strategies for patients with PDAC.
PMID:40848256 | DOI:10.1016/j.celrep.2025.116191
EClinicalMedicine. 2025 Aug 12;87:103406. doi: 10.1016/j.eclinm.2025.103406. eCollection 2025 Sep.
ABSTRACT
BACKGROUND: Recurrence risk after curative surgery for colorectal liver metastases (CRLM) remains high, underlining the need to identify prognostic markers enabling more individualised treatment approaches.
METHODS: In the MIRACLE, a prospective, observational biomarker study, a total of 188 patients with isolated, resectable CRLM without (neo)adjuvant chemotherapy were included between October 2015 and December 2021. Blood samples were collected before surgery (baseline) and three weeks after surgery. The primary objective was to assess the potential association between postoperative circulating tumour DNA (ctDNA) detection and recurrence of disease for patients with resectable CRLM within one year after resection. The secondary objective was the association between recurrence of disease within one year and detection of circulating tumour cells (CTCs). Baseline ctDNA was measured by next generation sequencing using a targeted panel (Oncomine Colon cell-free DNA assay) and postoperatively by digital PCR on genetic variants found preoperatively with the Oncomine panel. CTCs were enumerated using the FDA-approved CellSearch system.
FINDINGS: ctDNA was detected in 117/187 patients (63%) at baseline, and 28/104 evaluable patients (27%) still had detectable ctDNA postoperatively. CTC enumeration resulted in positivity for 37/183 patients (20%) at baseline and 14/158 patients (9%) postoperatively. No association was found between 1-year recurrence-free survival (RFS) and the presence of CTCs or ctDNA at baseline. In contrast, patients with postoperative undetectable ctDNA had a significantly improved 1-year RFS compared to patients with postoperative ctDNA (54% [95% CI 44%-67%] vs. 25% [95% CI 13%-47%], log-rank p = 0.0011). Similarly, patients with postoperative detectable CTCs had a significantly shorter 1-year RFS compared to patients without postoperative CTCs (15% [95% CI 4%-55%] vs. 53% [95% CI 45%-62%], log-rank p 0.0004). Also in multivariable analysis, detectable ctDNA and CTCs after surgery remained independently associated with a shorter 1-year RFS (HR 2.35; 95% CI 1.34-4.11; p = 0.0028 and HR 2.98; 95% CI 1.56-5.71; p = 0.0010, respectively).
INTERPRETATION: This is the first study conducted in patients with resectable CRLM without (neo)adjuvant chemotherapy, which demonstrates the impact of postoperative detectable circulating tumour load on 1-year RFS. Postoperative ctDNA and CTC detection both represent strong, independent predictors for a shorter RFS after local treatment, as opposed to preoperative detection.
FUNDING: This work was supported by KWF Kankerbestrijding (Dutch Cancer Society, EMCR 2014-6340).
PMID:40838198 | PMC:PMC12361997 | DOI:10.1016/j.eclinm.2025.103406
Ann Rheum Dis. 2025 Aug 20:S0003-4967(25)04249-9. doi: 10.1016/j.ard.2025.07.016. Online ahead of print.
ABSTRACT
OBJECTIVES: Catastrophic antiphospholipid syndrome (CAPS) is a complement-driven thrombotic disorder, characterised by widespread thrombosis and multiorgan failure. We identified rare germline variants including complement receptor 1 (CR1) in 50% of patients with CAPS. Here, we define CR1 dysregulation mechanisms (genetic/epigenetic) underlying complement-mediated thrombosis in CAPS and support C5 inhibition as a potential therapy.
METHODS: We quantified CR1 expression by flow cytometry across haematopoietic cell types. CRISPR/Cas9 genome editing of TF-1 (erythroleukaemia) cells was performed to generate CR1 'knock-out' and 'knock-in' lines with patient-specific CR1 variants. Multiomics analysis was performed to investigate the role of methylation in patients with reduced CR1 expression. Functional impact of low CR1 was assessed by complement-mediated cell killing using modified Ham assay, cell-bound complement degradation products through flow cytometry, and circulatory immune complexes in serum samples through ELISA.
RESULTS: CR1 expression in erythrocytes was markedly reduced on CAPS erythrocytes (n = 9, 21.80%) compared to healthy controls (HCs; n = 35, 84.04%), with promoter hypermethylation emerging as a plausible epigenetic mechanism for CR1 downregulation. Novel germline variant (CR1-V2125L; rs202148801) mitigated CR1 expression and increased complement-mediated cell death of knock-in cell lines. Erythrocytes from the patient with the CR1-V2125L variant had low CR1 expression. Levels of circulating immune complexes, which are bound and cleared by CR1 on erythrocytes, were higher in acute CAPS (n = 3, 25.55 µg Eq/mL) than HCs (n = 3, 7.445 µg Eq/mL). Five patients were treated with C5 inhibition which mitigated thrombosis.
CONCLUSIONS: Genetic or epigenetic-mediated CR1 deficiency is a potential hallmark of CAPS and predicts response to C5 inhibition.
PMID:40841298 | DOI:10.1016/j.ard.2025.07.016
Login