Could AI kill us all?

A possibly staged viral video from Russia recently showed a humanoid robot answering a customer’s push with punches and kicks. Imagine the Dhaka version: a customer enters a showroom, calls the robot “bekar,” pushes it, and receives an algorithmically precise slap. Within minutes, three agencies claim jurisdiction, two committees form, and the robot gets bail before the victim reaches police.

The joke becomes uncomfortable beside CNN’s interview with Jacob Coxon, a researcher who resigned from Anthropic after working there and at OpenAI. He says labs are racing towards self-improving superintelligence without solving “alignment”: keeping a machine more capable than its makers faithful to human intentions. Anthropic researcher Evan Hubinger estimated a greater than 10 percent chance of AI killing humanity within a decade. These forecasts are disputed. Yet when engineers building the aircraft question its brakes, passengers should stop arguing over the menu.

Tap here to add The Daily Star as a trusted source

The choice is between structured AI adoption and digital improvisation, with poor governance risking jobs, privacy and accountability

Bangladesh is unlikely to manufacture rebellious superintelligence. We may import it through an app, accept terms unread, and discover that our data, decisions and jobs travelled abroad. AI enables deepfakes, fraud, cyberattacks and misinformation. Where rumour outruns broadband and police, synthetic audio of a politician, religious leader or businessperson saying something incendiary could become a weapon.

Our policy response is promising on paper but unfinished. Bangladesh’s National Artificial Intelligence Policy 2026–2030 remains Draft V2.0; consultation ended in February. It proposes a risk-based framework, impact assessments, digital sovereignty and a Bangla language model. Personal Data Protection Ordinance and Cyber Security Ordinance provide architecture. But a draft cannot investigate an algorithm, compensate a victim, or switch off a dangerous system. Bangladesh has a tradition of policies whose implementation arrives by a later train, after the station has been renamed.

The gap is sobering. Oxford Insights placed Bangladesh 75th among 195 governments in its 2025 AI Readiness Index, up from 82nd, while India stood 27th and China eighth. Progress matters, but rank is not readiness. UNESCO highlights gaps in connectivity, computing capacity, skills, research, datasets and institutional coordination. Enforcement bodies struggle with yesterday’s offences; asking them to audit opaque algorithms tomorrow without people, laboratories or budgets would be science fiction.

Adoption could bring gains. Bangla AI tutors could reach students; diagnostic tools could support rural doctors; farmers could receive crop and weather advice; banks could detect fraud; factories could reduce waste; and government could target benefits and taxes. Exporters can raise productivity as buyers demand speed and traceability.

But unmanaged efficiency can become organised exclusion. Routine work in garments, outsourcing, banking, media and administration may shrink before retraining. Biased systems may deny credit or jobs, surveillance may become cheaper, foreign models may misunderstand dialects, and firms may capture gains while displaced workers receive motivational webinars.

Bangladesh needs more than another launch ceremony. Finalise the policy, enact enforceable risk rules, name one lead authority, require audits for high-risk systems, mandate human appeal, report serious incidents, protect children and biometric data, and create sandboxes for health, finance and public services. Invest in Bangla datasets, national compute, university research and reskilling.

The choice is not between worshipping AI and banning it. It is between structured adoption and digital improvisation. If we teach machines everything about us before teaching institutions how to govern them, the first blow may not come from a robot’s fist. It may arrive quietly, in a rejected loan, a fabricated speech, a vanished job or a decision for which nobody accepts responsibility.

The writer is founder of BuildCon Consultancies Ltd and BuildNation Ltd.