Office of Attorney General Bangladesh

For the Global South, the Dialogue is a success only if it shifts from abstract ethics to functional liability. Key outcomes include: Recognition of "Social Death": Formal acknowledgement that synthetic media harms in conservative societies constitute "Social Death"—the total annihilation of a person's communal standing—requiring higher legal protection than standard privacy breaches. A Shift in the Burden of Care: Moving from victim-led reporting to Strict Platform Liability. Success means global consensus that platforms must embed safety infrastructure (like localized "Take It Down" judicial links) before deploying tools in new markets. Decoupling Speech from Technical Harm: Distinguishing between regulated opinion and technical violence. Bangladesh's recent withdrawal of 410 "speech offense" cases serves as a blueprint: moving away from policing expression toward penalizing the technical weaponization of AI. Interoperability of Digital Justice: Establishing standards that allow national architectures, like Bangladesh's Cyber Security Ordinance 2026 and RPO 2026 amendments—to mandate platform cooperation across borders.AI governance that cannot see the harm cannot protect the people who suffer it. To be a success, the Global Dialogue must move beyond a one-size-fits-all approach and recognize that for the majority of the world's population, AI liability is not a technical debate, it is a matter of survival and dignity.

• Transparency, accountability, and human oversight
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Protection and promotion of human rights
• Interoperability of governance approaches

The following four areas from Resolution 79/325 as urgent for Bangladesh: Transparency, Accountability, and Human Oversight: This aligns with the "Digital Referee" model. AI must be auditable to identify where it fabricates or misinterprets evidence. Accountability must rest with the developer, not the end-user or victim. Protection and Promotion of Human Rights: AI governance is a gender justice emergency. In Bangladesh, 89% of women face online abuse. We prioritize frameworks that treat the right to digital dignity as a fundamental, non-negotiable human right. Social, Economic, Ethical, Cultural, Linguistic, and Technical Implications: AI is not culturally neutral. We must address the "Patriarchal Mirror"-where AI trained on Western or biased datasets exports harms to the Global South. Technical design must reflect regional socio-legal stakes. Interoperability of Governance Approaches: Our Cyber Security Ordinance 2026 requires a global bridge. Governance is only effective if local mandates for content removal or data security are enforceable against global tech giants.

The listed themes fail to capture the "Liability Void" affecting developing nations. Two critical issues emerge: Algorithmic Colonization: Global platforms profit from the data of the Global South but provide zero localized "Safety Engineering." While Western users have robust reporting tools, Bangladeshi victims face a linguistic and bureaucratic wall. We propose "Correctional Jurisprudence"-mandating that AI be designed to audit and neutralize regional prejudices at the source. The "Take It Down" Judicial Link: Emerging models like Thailand's 2026 fast-track judicial removal system represent the future of digital justice. We must bridge the gap between national judicial orders and algorithmic execution. In conservative contexts, a delay of 24 hours in removing a deepfake is the difference between a life preserved and a "Social Death" occurring. The Dialogue must address Extraterritorial Accountability: platforms must be held liable in the jurisdictions where they cause harm, regardless of where their servers are located.

Challenges: The Governance Gaps The "Social Death" Liability Void: A critical gap exists where global AI models, trained on Western datasets, are deployed in conservative social contexts without adequate safety guardrails. This results in "Social Death", the total destruction of a person's communal standing due to deepfakes, which existing Western-centric laws treat merely as privacy violations rather than existential harms. The Patriarchal Mirror in Technical Design: Current AI systems in the region often function as a "Patriarchal Mirror," perpetuating historical inequalities because women and marginalized groups are excluded from the data generation and technical design phases. Platform Accountability Gaps: While countries like Bangladesh are reforming local laws, there remains a massive enforcement gap where global tech platforms ignore regional legal mandates for content removal, leaving victims to bear the full burden of reporting and proof. Systemic Exclusion in Data: Decision-making institutions and data processes in Southeast Asia often lack diverse female voices, leading to AI assumptions that default to patriarchal biases. Opportunities: Developments and Advances Pioneering "Correctional Jurisprudence": Bangladesh is moving from passive bias mitigation to "Correctional Jurisprudence", an approach that treats AI as a "Digital Referee" designed to audit and correct human prejudice rather than institutionalize it. Legislative Decoupling of Speech and Crime: The dismissal of over 400 "speech offense" cases and the transition from the Digital Security Act to the Cyber Security Ordinance 2025/2026 represents a major advance. It signals a shift toward focusing on technical AI threats and data breaches while protecting free expression. Integrating AI into Electoral Integrity: The amendment of the Representation of the People Order (RPO) to ban deceptive AI content in election campaigns serves as a regional model for safeguarding democratic choice against algorithmic manipulation. Encoding Liberation: There is a transformative opportunity to treat gender parity as a normative indicator of institutional health, deliberately encoding inclusion into the legal and public administration systems of the state.

The AI Dialogue can bridge between the technical speed of the Global North and the socio-legal realities of the Global South. For a jurisdiction like Bangladesh, it represents an opportunity to transition from being a passive recipient of AI tools to an active architect of global standards. The Dialogue can establish an international interoperability framework that mandates extraterritorial accountability, ensuring that tech giants provide the same safety infrastructure in Dhaka or Bangkok as they do in Brussels. Standardizing "Correctional Jurisprudence", the Dialogue can facilitate this mandate and make a shift from passive bias mitigation to the active creation of equity. By sharing frameworks where AI acts as a "Digital Referee" to audit human prejudice, the international community can ensure that "responsible innovation" includes the proactive protection of marginalized groups from synthetic violence. By examining successful regional models like Thailand's 2026 "Take It Down" mechanism, the AI Dialogue can ensure technology transfer using technical assistance program for ensuring digital justice by standardizing how national judicial orders for content removal are communicated to and executed by global algorithmic systems.

The AI Dialogue is the only forum capable of ensuring that AI governance "sees the harm" unique to conservative social contexts, effectively transforming a global technical malfunction into a universal pillar of institutional integrity. The AI Dialogue can bridge between the technical speed of the Global North and the socio-legal realities of the Global South. For a jurisdiction like Bangladesh, it represents an opportunity to transition from being a passive recipient of AI tools to an active architect of global standards to ensure technology transfer using a technical assistance program for ensuring data preservation and digital justice by standardizing how national judicial orders for content removal are communicated to and executed by global algorithmic systems. The Dialogue can establish an international interoperability framework that mandates extraterritorial accountability, ensuring that tech giants provide the same safety infrastructure in Dhaka or Bangkok as they do in Brussels. Standardizing "Correctional Jurisprudence", the Dialogue can facilitate this mandate and make a shift from passive bias mitigation to the active creation of equity. By sharing frameworks where AI acts as a "Digital Referee" to audit human prejudice, the international community can ensure that "responsible innovation" includes the proactive protection of marginalized groups from synthetic violence. The Dialogue's unique value lies in its ability to provide Extraterritorial Enforcement Power. While national laws, like Bangladesh's Cyber Security Ordinance 2026, can mandate local safety, they often lack the leverage to compel global tech giants to comply. The AI Dialogue can, fast-track judicial remedy pioneered by Thailand to ensure deepfakes causing social harm are removed globally, not just locally. Play the role of a neutral arbitrator between national security needs and global tech profit motives, ensuring that "responsible innovation" is never prioritized over human dignity.

To advance a truly inclusive Global Dialogue, stakeholders must move beyond advisory roles into "co-architect" positions. Based on the framework of Correctional Jurisprudence, act like "Regulatory Lead Pilots," sharing localized legislative models like Bangladesh's Cyber Security Ordinance 2026 and RPO amendments to demonstrate how AI can be decoupled from speech offenses and focused on technical harms. Private Sector & developers must provide "Explainable AI" (XAI) access to national regulators, allowing for the auditing of patriarchal or regional biases in LLMs before deployment. Civil Society & Academia serve as "Digital Auditors," documenting real-world impacts of "Social Death" and ensuring the Dialogue prioritizes gender parity as a technical, not just ethical, requirement.

In global AI governance, the most dangerously underrepresented perspective is the "Global South User in Conservative Social Contexts." Current discussions are dominated by Western liberal assumptions that treat AI harm as a fixable technical deficit or a data privacy breach. For women in countries like Bangladesh, the primary threat is not "misinformation" but "Social Death." When synthetic media is weaponized for gender-based annihilation, the harm is existential and irreversible. Their voices are missing from the design of liability frameworks. AI models are often "linguistically colonizing," trained on Western datasets that fail to understand regional nuances, leading to high error rates in automated moderation for languages like Bengali. To Include Them governance frameworks should require AI developers to conduct safety testing with local civil society in the Global South to identify culturally specific harms before deployment. Move beyond "consultation" to "co-design." Stakeholders from underrepresented regions should help engineer Correctional Jurisprudence, designing AI that proactively audits for patriarchal and regional biases. As suggested by Resolution 79/325, the AI Dialogue must provide voluntary fund-supported participation for experts from developing countries to ensure that "international cooperation" is not a Western-centric luxury. Inclusion means moving from "designing for" the Global South to "designing with" the legal architects who actually navigate its unique challenges.

Instead of static speeches, use live simulation labs. Participants, including developers, legal experts, and civil society—would be presented with a real-world "safeguard failure" (e.g., a viral deepfake targeting a female candidate during an election). Stakeholders must collaborate in real-time to apply their respective frameworks, identifying where the "Liability Void" exists and how "Correctional Jurisprudence" could neutralize the harm. Create a format where legal architects from different jurisdictions (e.g., Bangladesh and Thailand) "cross-examine" each other's frameworks. For example, presenting Thailand's 2026 "Take It Down" mechanism to a panel of global tech representatives to stress-test technical interoperability. This moves the discussion from "if" we should regulate to "how" the court order reaches the algorithm. To bridge the empathy gap, use anonymized, immersive case studies of victims of "Social Death." These briefings would ensure that every policy discussion is grounded in the existential stakes of the Global South, shifting the metric of success from "innovation speed" to "human dignity preserved."

Effective AI governance is shifting from abstract ethical principles to solutions that address the technical risk and human justice of the 2026 digital landscape particularly in the Global South. One of the most significant breakthroughs in 2026 is Thailand's "Take It Down" mechanism, launched via the Court Integral Online Service (CIOS) which allows victims of AI-generated sexual harassment or "nudification" to petition the court electronically for an immediate takedown of illegal content and bypasses lengthy police investigations. In the first quarter of 2026 alone, the Thai judiciary ordered the removal of over 380,000 illegal URLs, demonstrating how digital infrastructure can scale to meet the volume of synthetic threats. This was preceded in the Philippines, Indonesia, and Malaysia temporarily restricting access to the AI tool Grok after it was used to generate non-consensual deepfakes. These nations demonstrated that if a tool poses an immediate risk to "online dignity" and the platform refuses to implement safety modifications, a suspension is a legitimate enforcement tool. For the same reason, India took a stricter stance by amending its IT rules, which made the platforms legally liable for the content as if they were the author if they fail to remove non-consensual deepfakes within 2 hours. Otherwise, they lose their "Safe Harbor" protection if they cannot comply. To combat the threat of deepfakes in the 2026 parliamentary elections, the Bangladesh Election Commission (EC) integrated, where deceptive synthetic content is treated as a form of electoral fraud, allowing for the disqualification of candidates who use AI to manipulate the "will of the people." Technical Sovereignty mandates that critical datasets remain within the country's sovereign control while ensuring they are formatted for Interoperability. Training AI models on localized, linguistically inclusive data prevents the "Patriarchal Mirror" effect, where imported AI fails to understand the cultural nuances of the Global South, leading to biased or harmful decisions.