The Presidential Secretariat has directed the Digital Economy Ministry to take forward what would be the world’s first national pilot of “universal basic AI”, according to Chamara Somaratne, founder of the Melbourne-based firms Anthosa and Sovereign Stack, the Daily Mirror reported.
Somaratne said the directive was issued this week, after two years of representations to President Anura Kumara Dissanayake — who holds the Digital Economy portfolio — and Senior Presidential Adviser on Digital Economy Dr. Hans Wijayasuriya. The pilot is to run in Sri Lanka and Australia.
“Now this is actually formally sanctioned by the president himself,” he said.
The claim rests on one speaker
Every element of this account, including the existence and contents of the directive, comes from Somaratne. Neither the Presidential Secretariat nor the Digital Economy Ministry has issued a statement, and no official confirmation of the directive had been published at the time of writing. He is also the commercial party that would build what he says has been approved.
The problem he describes
Somaratne puts Sri Lanka’s annual spending on foreign digital services at about US$600 million — over Rs. 180 billion — which he says exceeds a food import bill of roughly US$550 million. The outflow, on his account, spans cloud services, software platforms from Microsoft 365 to ride-hailing apps, and digital advertising. Australia, he said, spends around A$35 billion a year on US technology.
He described the pattern in blunt terms: “The way the digital economy in America is working at the moment is like a parasitic wasp. It’s actually eating the physical economy alive inside out.”
What is proposed
The framework centres on Sovereign Engram, an application installed directly on personal laptops that Somaratne says handles up to 80% of everyday AI work with no cloud dependency, no subscription and no data collection. It ships with a 4.5GB model, and he claims it reaches 140–160 tokens per second on a standard MacBook — five to seven times what he says open-weight models typically manage on the same hardware. These performance figures are the company’s own and have not been independently tested.
Heavier workloads would run on PicoDC nodes: low-power computing units on solar and battery storage, co-located at renewable sites, which he argues remove the case for hyperscale data centres. “We don’t need these massive data centres to be built,” he said.
Rollout is to begin on macOS, followed by Linux, with Windows targeted within two to three months. A Sri Lankan operating entity is being set up; initial PicoDC hardware would be imported, with local last-mile assembly planned. Somaratne said the company has submitted three open standards to the IETF, W3C and IEEE covering model packaging, inference and provenance verification.
