OpenAI mentioned Monday it might grant the European Union entry to its new cyber mannequin, however Anthropic continues to be holding out on releasing Mythos to the bloc.
European companions together with companies, governments, cyber authorities and EU establishments such because the EU AI workplace, can be granted entry to OpenAI’s GPT-5.5-Cyber, a variation of its newest AI mannequin, the corporate mentioned.
OpenAI introduced it was rolling out the mannequin in restricted preview capability to vetted cybersecurity groups. It got here a month on from Anthropic releasing its personal mannequin, Mythos, that prompted a wave of fears round cyberattacks on vital software program.
“We welcome OpenAI’s transparency and intent to provide fee entry to new mannequin,” Fee Spokesperson Thomas Regnier mentioned at a press briefing on Monday.
He confirmed an change had taken place between OpenAI and the EU, and that there have been additional discussions deliberate this week round entry to the mannequin. “This may enable us to observe deployment of the mannequin very carefully, and deal with safety considerations,” he added.
Although Mythos was launched a month in the past, Anthropic has but to grant the EU preview entry to evaluate it.
The EU is discussing entry with Anthropic, Regnier mentioned, however added that the discussions are at a “completely different stage” than they had been with OpenAI.
Whereas the Fee had had “4 or 5” conferences with Anthropic, Regnier mentioned that discussions with the corporate had been “not but on the identical stage as the answer now we have on the desk from OpenAI.”
“AI labs like ours should not be the only arbiters of cyber security as resilience is dependent upon trusted companions working collectively,” OpenAI’s Head of OpenAI for Nations George Osborne, mentioned in an announcement.
“The newest cyber AI capabilities ought to be obtainable for Europe’s many defenders, not simply the few, and we wish to assist make that occur,” he added.
“By way of the OpenAI EU Cyber Motion Plan, we’ll work with European policymakers, establishments and companies by democratizing entry to the defensive instruments that trusted actors can use to strengthen shared safety, help public security and replicate European priorities.”
Anthropic has been approached for remark.
