OpenAI scrambles to repair damage after rushed US military deal backlash
OpenAI is facing heavy criticism in the United States and abroad after signing a fast‑tracked contract with the US Department of Defense to deploy its AI models on classified military networks, a move that many users, employees, and civil society groups saw as opportunistic and careless. The agreement was concluded just hours after rival Anthropic was blacklisted as a “supply chain risk” and lost a 200 million dollar Pentagon contract for refusing to relax strict limits on surveillance and autonomous weapons, creating the perception that OpenAI had rushed to occupy the space left vacant by its competitor.
According to reports, tensions began when US authorities clashed with Anthropic over clauses that banned the use of its Claude model for mass domestic surveillance and required meaningful human oversight over autonomous weapons systems, following its alleged involvement in an operation in Venezuela that led to the capture of Nicolás Maduro and dozens of casualties. After Anthropic refused to weaken those safeguards, President Donald Trump ordered federal agencies to stop using its products and the Pentagon reclassified the company as a supply chain threat, a designation more commonly applied to foreign firms such as China’s Huawei and one that forced defense contractors to certify they no longer relied on Claude.
OpenAI quickly stepped in, announcing an agreement that would allow its models, including those powering ChatGPT, to be used across classified Pentagon systems under a broad commitment to serve “any lawful purpose,” while claiming it had preserved its own red lines on surveillance and lethal autonomy. The timing and wording, however, raised questions about whether the company had accepted looser constraints than Anthropic, especially since early versions of the contract did not explicitly forbid the use of commercially acquired data such as location histories, browsing records, or financial information purchased from data brokers to support US domestic surveillance operations.
The announcement triggered a wave of cancellations and deletions of ChatGPT accounts, with users denouncing what they saw as a betrayal of previous ethical commitments and an alignment with the military‑industrial complex, while alternative tools such as Anthropic’s Claude gained visibility as potential replacements. Inside OpenAI, employees joined colleagues at other tech firms to issue an open letter urging management to refuse Pentagon demands for mass domestic surveillance and autonomous targeting without human control, warning that the deal risked normalizing uses of AI that many staff considered unacceptable.
Facing mounting public and internal pressure, CEO Sam Altman acknowledged that the agreement had been handled poorly, describing it as “opportunistic and rushed” and conceding that communication around the deal had been a mistake. In a message posted on X, he announced amendments to the contract that now explicitly ban the use of OpenAI systems for domestic surveillance targeting US citizens and restrict access for certain intelligence agencies, while adding references to “commercially acquired data” to close loopholes that could have allowed the government to feed purchased personal information into AI‑driven monitoring programs.
Altman has also argued that OpenAI sought to secure conditions similar to those demanded by Anthropic, including some limits on the deployment of fully autonomous weapons, but admitted that the company has little direct control over how the Pentagon ultimately integrates its models into military workflows once technical access is granted. Critics say this episode lays bare the growing tension within the AI industry between lucrative government and defense contracts on one side and public promises about safety, transparency, and human rights on the other, with OpenAI now trying to rebuild trust while maintaining its newly won position as a strategic supplier to the US armed forces.
-
11:20
-
11:15
-
11:15
-
11:00
-
10:45
-
10:30
-
10:25
-
10:24
-
10:15
-
10:00
-
09:45
-
09:30
-
09:15
-
09:00
-
08:55
-
08:45
-
08:40
-
08:30
-
08:25
-
08:15
-
08:00
-
07:45
-
07:30
-
07:15
-
23:57
-
23:42
-
23:34
-
23:23
-
23:15
-
23:15
-
23:00
-
22:57
-
22:42
-
22:22
-
22:10
-
22:00
-
21:40
-
21:30
-
21:21
-
21:20
-
21:15
-
21:00
-
20:45
-
20:30
-
20:20
-
20:15
-
20:00
-
19:45
-
19:30
-
19:15
-
19:00
-
18:45
-
18:30
-
18:15
-
18:00
-
17:45
-
17:30
-
17:15
-
17:10
-
17:02
-
17:00
-
16:45
-
16:30
-
16:15
-
16:12
-
15:55
-
15:54
-
15:39
-
15:20
-
15:20
-
15:08
-
15:05
-
14:45
-
14:42
-
14:30
-
14:14
-
14:08
-
14:00
-
13:55
-
13:35
-
13:20
-
13:05
-
12:45
-
12:30
-
12:15
-
12:15
-
12:00
-
11:42
-
11:37
