Anthropic Adds Security Measure to Mend Ties With Trump Administration

▼ Summary
– The Trump administration lifted export controls on Anthropic’s Claude Fable 5 AI model after the company agreed to extend a guardrail that blocks certain user requests and routes them to the less-advanced Opus 4.8 model.
– The new safeguard extends an existing restriction to cover a specific behavior identified in an Amazon paper, where users bypassed security limits by asking the model to fix code instead of identifying security issues.
– Commerce Secretary Howard Lutnick announced the removal of restrictions, stating Anthropic agreed to proactively detect and address security risks posed by the models.
– The Commerce Department cleared Fable 5 after its Center for AI Standards and Innovation deemed the safeguards sufficiently robust.
– Defense Secretary Pete Hegseth has told advisers there is no clear path to lift his February 28 order designating Anthropic a supply chain risk, leaving some challenges unresolved.
The Trump administration has lifted export controls on Anthropic’s Claude Fable 5 AI model after the company agreed to expand an existing security guardrail designed to block users from attempting to access specific restricted capabilities, according to two sources familiar with the negotiations.
Under the updated safeguard, any user who tries to unlock those capabilities will see their request denied and will receive a notification that their query has been rerouted to the less advanced Opus 4.8 AI model, the sources said.
Previously, Anthropic routed user requests related to sensitive cybersecurity and biology capabilities to Opus 4.8 before cutting off access to Fable 5 entirely. The new measure extends that guardrail to requests tied to a specific behavior highlighted in a paper published by Amazon.
An analysis by Katie Moussouris, founder and CEO of Luta Security, found that after reading the Amazon paper, users could bypass a restriction on Fable 5 by asking the model to fix code rather than identify security flaws in it. While most cybersecurity experts do not consider this behavior alarming, the administration’s awareness of it triggered a confrontation with Anthropic and led to the imposition of export controls, which effectively took the model offline.
The additional detail sheds light on Commerce Secretary Howard Lutnick’s letter announcing the removal of restrictions on Anthropic’s Fable 5 and Mythos 5 AI models. “Among other things, Anthropic has agreed to proactively detect and address security risks posed by the models,” Lutnick wrote, noting his role in bringing the models back online. WIRED first obtained the letter and shared its contents Tuesday night.
The Commerce Department ultimately cleared Fable 5 for release after researchers at its Center for AI Standards and Innovation determined the model’s safeguards were sufficiently robust for now, the sources added.
However, while Anthropic has resolved its standoff with the Commerce Department, Defense Secretary Pete Hegseth has told advisers there is no clear path to lift his February 28 order designating the company a supply chain risk, according to a person briefed on the matter.
So although some of Anthropic’s challenges with the administration have eased, they are not entirely behind them.
In a separate development, Trump administration officials believe the U. S. Supreme Court has handed them a key advantage ahead of the midterms. In a 6-3 decision on Tuesday, the court opened the door for the first time for political parties to coordinate messaging and spending with campaigns, paving the way for Republicans to leverage Trump’s massive fundraising operation.
In the short term, administration officials tell Inner Loop that the ruling in National Republican Senatorial Committee v. Federal Election Commission disproportionately benefits vulnerable Republican candidates in this midterm cycle.
Before the decision, individual Democratic candidates typically raised more money from small-dollar donors than their Republican counterparts, who often relied more on the Republican National Committee, funded by billionaires. A rule allowing candidates to qualify for cheaper television advertising rates also gave Democratic candidates more bang for their buck, though the effectiveness of TV ads is increasingly seen as declining.
Now, with national parties like the RNC permitted to buy unlimited ads on behalf of candidates at those lower rates while directly coordinating messaging and attack lines, the television advertising advantage may have shifted.
The Republican National Committee entered June with $125.5 million cash on hand and no debts. In contrast, the Democratic National Committee reported $14.9 million cash on hand with $18.3 million in debts, according to its latest filings from June.
Republicans also hold a cash advantage in the House and Senate campaign arms. The National Republican Congressional Committee and the National Republican Senatorial Committee each reported approximately $10 million more than their Democratic counterparts.
With November’s midterm elections approaching, this decision could not have come at a worse time for Democrats.
(Source: Wired)




