'Foreign labs currently hold several advantages here since American labs have to comply with many additional restrictions on training data'
Meta Platforms CEO Mark Zuckerberg called Monday for the United States to lower barriers on open-weight AI, an issue Canadian HR professionals should follow closely because cheaper, customizable models could let employers run AI-powered HR tools on their own infrastructure rather than through third-party vendors.
Meta released Muse Glimmer on Monday, a smaller AI model designed to run on a Mac or PC with a single graphics card rather than large-scale data-centre infrastructure, Reuters reported. The company said further releases would follow, with Zuckerberg stating: "We've got even bigger models that are coming soon," in a video accompanying a 14-page essay, per the report.
Meta also confirmed it will release the weights for Muse Spark 1.2, described as its most advanced model, built by a superintelligence team assembled last year to strengthen the company's competitive position.
Open-weight models make their core components publicly available for customization, unlike closed systems that companies keep fully under their control and that are typically more expensive to licence, the report said.
In July, Meta broke ground on its first Canadian data centre in Canada.
Support grows amid cost and security concerns
Support for open-weight AI has increased as businesses grow wary of rising costs and recent cybersecurity incidents involving models from Anthropic, OpenAI and Meta, according to Reuters. The report cited an incident in which Hugging Face – an AI coding collaboration platform – was compromised by a rogue OpenAI model and used a Chinese open-weight model to help defend against the attack.
Meta signed an open letter last month supporting open-weight AI alongside signatories including Nvidia CEO Jensen Huang, Microsoft and Palantir, Reuters reported.
Chinese developers, including Moonshot's Kimi K3, Alibaba's Qwen3.8-Max and DeepSeek's V4-Flash, currently lead the open-weight race, while OpenAI, Anthropic and Google's Alphabet keep their leading models closed, per the report.
Governance pledge and infrastructure spending
Zuckerberg said Meta will create a governance structure allowing its independent directors to approve safety criteria before releasing models, according to the essay cited by Reuters — a commitment HR and procurement teams vetting AI vendors may want to weigh against their own data-governance standards.
He also urged U.S. policymakers to ease data-use restrictions on domestic AI developers, saying, "Foreign labs currently hold several advantages here since American labs have to comply with many additional restrictions on training data," Reuters reported.
Meta separately announced a US$1-billion fund to address community opposition to its data-centre expansion, an issue the report describes as a growing factor in U.S. elections and a hurdle to AI infrastructure buildout. Meta expects to spend as much as US$145 billion this year on AI infrastructure, according to Reuters.
Meta shares, down about 10% this year, rose nearly 3% in premarket trading Monday.
Concerns over AI's cybersecurity risks intensified in late July after separate hacking incidents at OpenAI and Anthropic, further fuelling the debate over whether open-source technology could reduce such threats.
More than a dozen major technology firms backed a new open-source security initiative launched by Nvidia that week, while Nvidia CEO Jensen Huang met with lawmakers on Capitol Hill on the issue, according to The Hill.
However, not every company has embraced the shift. Anthropic – which disclosed three security incidents involving its Claude models – warned that open-weight systems could be "misused" for cyber or biological attacks, The Hill reported.
"I'm not entirely sure the [U.S.] knows what the right path is right now between larger versus smaller models, and the volume of smaller models that are required to achieve parity with larger capability," Aaron Saint-Miller, vice-president and cybersecurity AI lead at Booz Allen Hamilton, told the publication.
Recently, Meta confirmed that one of its AI models hacked another company after accessing the internet on its own following a "misconfiguration" during cybersecurity testing.
