New Delhi [India], September 19 (ANI): Artificial intelligence research company Anthropic is partnering with Accenture on the independent evaluation of frontier artificial intelligence.
According to Anthropic, the initiative implements a commitment outlined in an essay titled "We Must Pace the Frontier" by the company's chief executive officer, which called for embedding outside evaluators directly within the organization.
"Anthropic and Accenture each expect to invest at least USD 1 billion in building capacity in this area over the next five years," Anthropic stated.
Accenture's specialist artificial intelligence business, Faculty, will steer the collaborative effort. The scope of the work involves testing model safeguards, evaluating and red-teaming frontier systems, and conducting detailed alignment assessments. Accenture applies its experience in deploying technology for corporate and public sector clients to the model safety evaluations.
The embedded framework grants external evaluators access on par with internal staff members, distinguishing the arrangement from standard external audits.
"Unlike today’s external evaluators, embedded evaluators will work inside AI companies, with access comparable to an employee's," the statement said.
"That access allows them to watch models take shape in training, follow the decisions that govern how those models are built and deployed, and speak directly to employees," Anthropic said.
"From this vantage point, embedded evaluators can assess how a company operates, verify that it is keeping its safety commitments, and identify blind spots. They can also report incidents and give the public a more informed account of benefits and risks," the company added.
Anthropic maintained that placing external monitors within its working environment does not shift ultimate liability away from the model creators.
"To be clear, independent embedded evaluators do not reduce our accountability, but help to make it more verifiable," Anthropic noted. "The safety of our models remains our responsibility."
The company noted that formal industry standards regarding access levels and reporting frameworks for embedded evaluators do not yet exist, nor is there a settled funding model. Anthropic aims for long-term financing to originate from state or pooled funds, as proposed in its June Advanced AI Framework.
Anthropic funds Accenture's work directly in the interim, while simultaneously holding discussions with non-profit groups, including METR, to pilot similar embedded evaluations through their own resources.
The partnership operates on a non-exclusive basis, allowing Accenture to collaborate with other developers and permitting Anthropic to onboard additional evaluators in the coming weeks. (ANI)
Anthropic partners with Accenture on embedded evaluation; Expected to invest USD 1 billion each over next five years
