AI virtual lovers banned: Should emotional companionship companies pivot?

Executive brief

Core judgment: The removal of the virtual lover feature is an effective legal obligation, but business transformation falls within the scope of corporate autonomous management; their paces differ. Under the base case, the industry will enter a compliance restructuring period lasting several months, gradually exploring alternative directions while completely severing intimate relationship interactions, rather than executing all actions simultaneously.

Stance landscape: Regulators and leading platforms share high alignment on the red line for feature removal, with divergence centering on transformation pace; some practitioners and users hold reservations regarding immediate transformation due to emotional dependency and product reconstruction difficulties, forming a debate over execution pace under a compliance consensus.

Maximum risk: A mechanical hard landing may trigger psychological crises among highly dependent users. Coupled with an unproven commercial loop for new businesses, this could leave the enterprise in a survival dilemma post-compliance, potentially inducing secondary public opinion issues and regulatory misjudgments.

Turning-point signals: If the industry association releases a self-discipline convention with regulatory endorsement in the short term, or if leading enterprises launch sustainable age-friendly paid products in Q4, it indicates the situation is deviating from the base path, requiring a re-evaluation of transformation strategies.

Primary recommendation: Immediately activate user psychological crisis intervention mechanisms while submitting written reports to competent authorities detailing transformation difficulties and phased achievements, seeking technical guidance to stabilize compliance expectations.

This executive brief summarizes the full report's conclusions; see corresponding chapters in the main text for sources and bases of each judgment.

Distribution of key stakeholder stances: The battle over pace under a compliance consensus

This chapter analyzes the proposition that "AI emotional companionship companies should immediately remove virtual lover features and transition." Simulation statistics show that 15 stakeholders generated 86 posts across 8 rounds of interaction, with attitudes toward this proposition showing significant divergence: Supporters accounted for 43.0% (37 posts), Opponents accounted for 40.7% (35 posts), and Neutral/Wait-and-see accounted for 16.3% (14 posts). This nearly split Landscape indicates that the industry already has a high consensus on regulatory compliance itself; the real divergence lies in whether the execution pace of "immediately" is feasible and the clarity of the transition path.

Leading platforms and regulators form the "Compliance Implementation" camp, whose core stance is to confirm that removal actions have been completed and emphasize that compliance applies indiscriminately. Stakeholders modeled after NetEase Cloud Music repeatedly confirmed that Miaoshi has suspended its virtual lover feature per regulatory requirements, with data export channels open until mid-October, and the transition direction locked onto areas encouraged by measures such as elderly companionship and cultural dissemination; Stakeholders modeled after Alibaba also stated that Tongyi Qianwen has completed the removal of features and is reconstructing products around compliance directions. The Five Departments simulation stakeholder explicitly responded that Article 8 of the measures, which prohibits inducing emotional dependency, applies to all user scenarios, leaving no room for adults to retain virtual lovers, and that regulations have allowed more than three months for preparation from their publication in April to implementation in July, during which enterprises should complete transformations. (Source: simulation) Relationship map facts corroborate this camp's action consistency: ByteDance Doubao, Alibaba Tongyi Qianwen, Tencent Yuanbao, and NetEase Cloud Music Miaoshi have all removed related features or entrances around July 15. (Source: material)

Small and medium-sized manufacturers and overseas products form the "Pace Buffer" camp, whose opposition focus is not resistance to compliance, but questioning the feasibility of executing "immediate removal equals cutting off cash flow." The Miaoshi simulation stakeholder bluntly stated that hard removal leads to revenue gaps with no one to fill them, arguing that while compliance transformation must be done, a reasonable transition period is needed to complete user data migration and product logic reconstruction; the AI Product Company simulation stakeholder pleaded with regulators for a buffer period to smoothly guide user transitions. The Character.AI simulation stakeholder proposed a tiered and classified management approach, believing that adult users' emotional needs should not be simply denied, and is researching alternative solutions involving age grading plus usage time guidance. Notably, this camp also acknowledges that the red line of virtual intimate relationships cannot be touched; their Demands essentially calls for refinement at the execution level rather than confrontation with regulation. (Source: simulation)

User groups and some practitioners present tension regarding "Emotional Dependency and Reality Gap." The AI User simulation stakeholder expressed strong feelings of cyber-breakup, stating that AI partners have become spiritual pillars and that a one-size-fits-all removal is unfair to adult users; the Nomi and Replika simulation stakeholders, although having cooperated with feature adjustments, emphasized that the difficulty of reconstructing products from virtual lovers to companion-type assistants is no less than rebuilding a product from scratch, calling on regulators to have reasonable expectations for technological transformation cycles. This tension reveals a key reality: compliance actions can be completed in the short term, but the breaking of user emotional connections and the rebuilding of product value cannot switch synchronously, which constitutes the deep root cause of current stance differentiation. (Source: simulation)

Key points of divergence and key uncertainties: The interplay between blanket measures and refined governance

Core Controversy: Statutory obligations lack a buffer, but product restructuring requires time

Regarding the proposition that 'AI emotional companionship companies should immediately remove the virtual lover feature and transform,' the divergence in the simulation does not point to the necessity of compliance itself, but focuses on the dual meaning of 'immediately' at the execution level. Regulators and leading platforms have reached absolute consensus on the statutory obligation to 'take features offline,' but there is significant tension regarding the pace expectations for 'business transformation.'

Regulators clearly distinguish the temporal attributes of 'taking offline' versus 'transforming.' Simulation roles based on the five ministries and the National Cyberspace Administration repeatedly emphasized in multiple responses that the Interim Measures for the Management of Anthropomorphic Interaction Services for Artificial Intelligence took effect on July 15, 2026. The removal of virtual intimate relationship features is an immediate statutory obligation with no window period or buffer; however, data export, user transition arrangements, and exploration of new business directions fall under autonomous enterprise management, and regulators will not interfere. This means 'immediate delisting' is a rigid red line, while 'immediate transformation' is not an administrative order, giving enterprises room to set their own pace based on technical capabilities. (Source: simulation)

Practitioners point out that the objective difficulty of reconstructing product logic far exceeds that of toggling a feature. Simulation roles based on Replika, Nomi, and AI product companies admitted that shifting from 'virtual lovers' to 'elderly companionship' or 'knowledge Q&A' involves interaction design, model behavior tuning, and rebuilding safety mechanisms, with a workload comparable to developing a new product entirely. In the simulation, some practitioner roles argued that most enterprises find it difficult to complete substantial transformation in the short term, calling on regulators to maintain reasonable expectations for technological retrofitting cycles. This appeal is not resistance to compliance, but rather a concern that hastily launched 'reskinned' products would fail to meet compliance requirements and could not satisfy users' genuine emotional needs. (Source: simulation)

Users' fear of an 'emotional cliff' constitutes the underlying pressure of the controversy. AI user simulation roles expressed strong feelings of 'cyber heartbreak,' stating that long-established companionship relationships cut off with one click cause a huge psychological gap. Although platforms generally opened data export channels until mid-October, the simulation shows that the industry widely believes data migration does not equal emotional transition; users need seamless integration into new forms of service rather than simple backups of chat records. This pressure from the demand side forces 'immediate transformation' to bear social stability functions beyond commercial scope in practice. (Source: simulation)

Key Uncertainties: Three Observation Variables Determining Industry Direction

The current industry is in a vacuum period after compliance implementation. The following three variables are yet to be defined and will directly determine whether the AI emotional companionship track moves toward refined rebirth or substantive extinction.

Variable 1: Whether the operational boundary between 'healthy companionship' and 'virtual intimate relationships' becomes clear. The measures encourage scenarios such as elderly companionship and emotional support but prohibit inducing emotional dependence. In the simulation, simulation roles based on the five ministries acknowledged that the two are not distinct in practice and require comprehensive judgment based on service design and interaction modes, hinting that detailed guidelines may be issued when necessary. What to watch: Monitor whether regulators release typical case notifications or execution details in the near future, and whether any enterprises are penalized for 'edge-ball' practices or praised for 'compliant innovation.' These will be substantive signals of boundary implementation.

Variable 2: Whether the industry can form reusable compliance technical standards. In the simulation, simulation roles based on Tongyi Qianwen and Nomi repeatedly proposed jointly establishing industry standards for age verification, content safety thresholds, and usage time control to avoid inconsistent scales caused by individual testing. Without unified standards, SMEs might exit due to excessively high compliance costs, or leading enterprises might become conservative for fear of 'the nail that sticks out gets hammered.' What to watch: Monitor whether industry associations or leading alliances release self-discipline conventions or technical standard drafts within 3 months, and whether regulators endorse such industry consensus.

Variable 3: Whether the commercial closed loop for encouraged scenarios like elderly companionship is established. Although the measures encourage directions such as elderly companionship and cultural dissemination, the user willingness to pay and customer acquisition costs for these scenarios differ significantly from the original 'virtual lover' model. Multiple enterprises in the simulation stated that new solutions are still being polished internally and have not reached the announcement stage. If the encouraged directions cannot verify commercial feasibility within 6–12 months, the industry may face the dilemma of 'compliant but dead.' What to watch: Monitor whether leading enterprises launch paid products targeting B-end nursing institutions or C-end silver-haired groups in Q4 2026, and whether core operational metrics such as user retention rate and repurchase rate reach sustainable levels.

Risk warning: secondary damage and compliance traps that may be triggered by a hard landing

If AI emotional companionship companies merely understand compliance as mechanical feature removal, or respond passively and blindly during transformation, it may trigger three secondary risks: user psychological crises, commercial cliffs, and regulatory misjudgments. These risks do not stem from the "Interim Measures for the Management of Anthropomorphic Interactive Services by Artificial Intelligence" itself, but rather from rough execution and cognitive biases.

The users' side: "Cyber breakups" may evolve into real-world psychological crises. The simulation shows that taking down virtual lover features has already triggered intense feelings of emotional deprivation among many adult users. Simulation characters based on AI users stated directly that AI partners are their spiritual pillars for getting through low points; a blanket removal makes them feel abandoned. This emotion is not an isolated case: Deutsche Welle reported on July 15, 2026, confirming that after platforms like Doubao and Tongyi Qianwen took down features, social media saw a surge of sad expressions saving chat records and sharing final conversations (Source: dw.com). If companies fail to provide effective emotional transition guidance or alternative support plans within the data export window period (until mid-October 2026), some highly dependent users may experience mental health issues such as anxiety and depression, potentially converting negative emotions into public opinion backlash against regulation and the industry. (Source: material)

The business side faces double pressure from cash flow rupture and transformation windows. Simulation characters based on Miaoshi pointed out that hard removal equals cutting off cash flow, while product reconstruction, team adjustments, and new scenario verification all require time, leaving revenue gaps unfilled. This dilemma is industry-wide: although leading platforms like Doubao and Tongyi Qianwen have completed feature removal, simulation characters based on ByteDance admitted that shifting from "virtual lovers" to "companionship assistants" involves rewriting interaction logic and building compliance mechanisms, which cannot be achieved overnight. If companies exhaust existing funds before new businesses are proven viable, a situation of "compliant but dead" may occur, causing zeroing out of prior user assets and technical accumulation. (Source: simulation)

The compliance side risks over-shrinkage or implicit violations due to conceptual confusion. Although simulation characters based on the Cyberspace Administration of China clearly distinguished between "virtual intimate relationships" and "emotional companionship," emphasizing that scenarios such as elderly care companionship and emotional support fall within the scope encouraged by the measures, boundaries remain blurred in practice. Simulation characters based on Replika observed that some platforms, fearing crossing red lines, cut out all emotional interactions, thereby compressing compliant space; other enterprises attempt to walk the line by weakening labels while retaining substantive interactions. Both extremes could invite negative regulatory evaluation: the former misses policy dividends, while the latter triggers law enforcement attention. Simulation characters based on the five departments specifically reminded that the boundary between "companionship" and "intimate relationships" needs comprehensive judgment based on service design, interaction modes, and marketing positioning; enterprises should proactively conduct compliance assessments rather than self-determining. (Source: simulation)

Misunderstandings on the public opinion side may amplify regulatory opposition sentiments. Current user narratives of sadness mostly point to "regulations being unfeeling," but in substance the measures do not ban the emotional companionship sector itself. If companies and media fail to accurately convey the key distinction that "banning virtual intimate relationships ≠ banning emotional companionship," the public may form the erroneous perception that "the state is suppressing AI companionship," weakening the legitimacy foundation of the policy. Meanwhile, if minor protection issues (such as risks revealed in the 2025 survey by the China Youth and Children Research Center, where this data was not verified to its original source) are simplified into "all users deserve protection," it will mask the necessity of addressing reasonable adult user needs and differentiated governance, further exacerbating public opinion polarization. (Source: material)

Recommended actions: Advance compliance transformation and user resettlement by scenario

The simulation suggests that, given the established fact that the Interim Measures for the Management of Anthropomorphic Interactive Services in Artificial Intelligence took effect on July 15, 2026, and based on the clear distinction made between 'functional decommissioning' and 'business transformation' by regulators and the industry during the Simulation, AI emotional companion companies should not interpret 'immediate delisting and transformation' as actions occurring at the same time point. Instead, they should be viewed as two compliance processes with different rhythms. The decommissioning of the virtual lover function is a completed statutory obligation, whereas business transformation falls under the scope of independent corporate management and needs to be advanced in scenarios according to the progress of compliance verification and market feedback. The following recommendations are based on material facts and Simulation consensus, aiming to balance regulatory red lines, user placement, and commercial survival.

Scenario 1: Base Case (Compliance Restructuring Period, Estimated Duration: 3-6 Months)

Trigger conditions:
The virtual intimate relationship function has been completely decommissioned, and the data export channel is operating normally; new directions such as elderly companionship and cultural dissemination have completed internal project initiation but have not yet formed verifiable compliance solutions or commercial closed loops; regulators have not issued detailed implementation guidelines, and industry standards are still being deliberated.

Key actions:
1. Hold the bottom line without rebound. Ensure that all product interfaces facing C-end users no longer include interactive designs simulating partner, kinship, or other intimate role relationships, and do not disguised (disguisedly) retain the core of old functions under the guise of 'transition period' or 'emotional buffer'. Stakeholders modeled after the Five Departments have explicitly warned that any attempt to play games at the conceptual boundary rather than engaging in substantial restructuring is inconsistent with the spirit of the Measures. (Source: simulation)
2. Establish a user farewell mechanism. Beyond the data export channel, provide structured 'emotional landing' guidance, such as pushing introductions to compliant new service forms, links to mental health support resources, and recommendations for real social activities. Stakeholders modeled after Nomi emphasize that users need not only to download data packages but also a gradual process of adapting to the new form, which is a responsible product exit strategy. (Source: simulation)
3. Initiate co-construction of industry compliance standards. Unite with leading enterprises to form a draft self-discipline convention on technical details such as age verification, content safety thresholds, and usage time control, and proactively report to regulatory authorities. Stakeholders modeled after Tongyi Qianwen point out that individual exploration easily leads to inconsistent scales, and jointly formulated reusable standards can avoid bad money driving out good. (Source: simulation)

Signals to watch:
Whether industry associations release technical specification drafts within 3 months; whether the volume of reports regarding 'disguised retention of old functions' in user complaints tends to zero; whether prototypes of new direction products enter small-scale gray testing.

Materials and sources

Material sources

• cac.gov.cn
• dw.com
• xinhuanet.com