A U.S. government agency is Anthropic’s Mythos to audit government software despite the Trump administration’s recent feud with the artificial intelligence company.
The Cybersecurity and Infrastructure Security Agency (CISA) is using Mythos to scan code in government systems that may be leaving doors open for foreign spies and cybercriminals, according to Reuters who cited three anonymous sources.
CISA, which formed in 2018, came under scrutiny after it became a tool for cracking down on “misinformation” through its facilitation of the Election Integrity Partnership. This group that worked closely with the agency flagged content to be removed from various social media sites during the 2020 election.
The scanning is carried out by CISA’s Attack Surface Evaluation team which conducts digital security assessments and hacking exercises across the government. The audits have already uncovered a large number of vulnerabilities but did not elaborate, the outlet reported, citing the anonymous sources.
Anthropic and CISA did not immediately respond to the Daily Caller News Foundation’s request for comments. CISA is an agency of the Department of Homeland Security (DHS).
Reuters was not able to establish how much of the government code the team had gone through or how serious the vulnerabilities were, the outlet reported.
“Given how the Biden administration has perverted the purpose of CISA to collude with Big Tech to establish a domestic censorship regime, the next Congress will need to reform the existing program, reassign authorities, and provide greater oversight, while holding accountable all government officials who used this government power improperly to target Americans with disinformation campaigns,” now-FBI Director Kash Patel told the Daily Caller regarding CISA in 2022.
Anthropic and the White House have been going back and forth on the use of Claude Mythos 5 and Fable 5.
In June the government ordered for Claude Mythos 5 and Fable 5 to be shut down from public access just three days after its release due to national security concerns. After Anthropic refused to allow its AI models to be given broad access to the U.S. military, President Donald Trump ordered for government agencies to immediately cease to use of its AI models.
“THE UNITED STATES OF AMERICA WILL NEVER ALLOW A RADICAL LEFT, WOKE COMPANY TO DICTATE HOW OUR GREAT MILITARY FIGHTS AND WINS WARS,” Trump said in a Truth Social post on Feb. 27.
At the beginning of Anthropic’s release the company admitted that security threats were a big possibility after its Mythos model escaped its “sandbox” testing environment.
Despite the feud the government lifted the ban in late June allowing the company to widely release the AI models after Anthropic pledged that they have stronger safeguards and closer cooperation with the government on future model releases.
Fable 5 is the general use version of Anthropic’s models and Mythos 5 is a more advanced model that is designed for high-level tasks with fewer built-in restrictions.
Axios reported that the National Security Agency has been using Mythos as far back as April despite the blacklist, according to Reuters. Anthropic has also confidentially filed for an initial public offering
(DCNF)
