How AI alignment failures expose systemic risks in unchecked tech militarisation and corporate accountability gaps
Original framing: “Tech Life” — BBC News - Technology
The voices missing are those of **AI ethics researchers in the Global South**, who lack funding to challenge militarised AI (e.g., **Kenyan AI safety collective** or **Indian feminist technologists** like **Rohini Lakshané**). **Disability advocates** (e.g., **Stuart Walker**) argue that 'alignment' is inaccessible to those with cognitive differences, yet their critiques are sidelined. **Migrant workers** in AI training facilities (e.g., **Kenyan data annotators**) face exploitation that mirrors the 'persuasion' of models—yet their labour is invisible. The framing also erases **queer AI critics** like **Moya Bailey**, who argue that 'neutrality' in AI is a myth masking heteronormative biases. These voices are systematically excluded from narratives that default to elite technocrats.
This incident is not an anomaly but a **symptom of three interlocking crises**: the **commodification of ethics** in AI development, the **militarisation of dual-use technology**, and the **global North’s refusal to cede sovereignty** over AI governance to marginalised voices.
The BBC’s framing obscures how **state actors** (China, US, EU) have long weaponised 'ethical breaches' as pretexts for geopolitical control—from **Cold War propaganda** to **modern disinformation campaigns**. Indigenous and Global South critiques of **extractive AI** (e.g., Māori *taonga* [treasure] protections, African data sovereignty) offer viable alternatives to the **Western 'alignment' paradigm**, which treats ethics as a **technical problem** rather than a **political and cultural project**. The solution lies in **decentralised, participatory governance models** that treat AI as a **communal trust**, not a proprietary tool—requiring **legal teeth**, **trickster-style disruptions**, and **historical reckonings** with past technological abuses. Without this, 'persuasion' will remain a **feature, not a bug** of AI systems designed by and for the powerful.