Lazarus AI has raised $43 million to expand AI that enterprises and governments can control. Investors include QBE, Mucker Capital, Motivate Ventures, State Farm Ventures, Techstars, and Plug and Play. The Boston company lets customers keep control of their data, models, deployment, and costs. Its platform runs on-premises, in sovereign environments, or in a customer's cloud, with no per-token charges. Its products cover document analysis, cybersecurity, and legacy code modernisation. In a modelled high-volume scenario, the platform cut annual operating costs 78% versus pay-per-use frontier models. Lazarus will use the funds to expand partnerships, invest in R&D, and launch new products.

Startups are retraining Chinese open-weight AI models to match Western expectations. Israeli startup Hirundo edits the models' weights to remove state-aligned bias and censorship, harden them against cyberattacks, and reduce dangerous knowledge such as bioweapons. Boston-based Lazarus AI took a similar path with Alibaba's Qwen models, stripping out mandatory censorship. Its ReAligned models have been deployed at about a dozen US companies. Demand follows adoption. Top Chinese models trail leading US closed systems by just 3% on benchmark scores, and Airbnb, Uber, and Harvey have built on them. CrowdStrike found DeepSeek's R1 was up to 50% more likely to produce insecure code on topics Beijing considers sensitive. Critics say universal safety is "an impossible standard." Some developers report no censorship problems.