Trump last week banned the three outlets from the White House, before a judge ordered the administration to restore their access.

The White House television press pool aired video of Xi’s arrival at the White House for tea on the final day of the Chinese leader’s visit, as well as the two leaders’ tour of the National Archives, which houses the US Constitution.

CNN, ABC, CBS, Fox News and NBC share the costs of covering the president and rotate coverage, providing pool footage to other outlets. The other networks’ refusal to set up cameras in solidarity meant there was no pool coverage of the first two days of Xi’s visit.

CNN on Friday evening said it had been barred from traveling with Trump on Air Force One for a weekend trip to Tennessee. The White House did not immediately respond to a request for comment from Reuters.

Trump has long relied on the media to shape his public image and advance his agenda, even as he regularly casts news organisations as political adversaries. Since returning to office last year, his administration has taken an aggressive stance against major news outlets, including limiting access to the Oval Office and other restricted spaces.

As pool coverage resumed for Xi’s arrival at the White House on Friday, Trump declined to answer questions from reporters on China and Iran, his media ban and a controversial new taxpayer-funded TV ad that promotes his presidency. He later thanked reporters.

“We appreciate the fair coverage. It’s been very fair. We appreciate it,” Trump later told reporters alongside Xi.

The White House did not respond to a request for comment about the restored press access.

Unprecedented campaign

Trump banned the three news outlets from the White House complex over coverage he asserted was false or negative. Their access was reinstated on Thursday, after a federal judge deemed the ban likely unconstitutional and ordered their access to be immediately restored.

CNN, MS NOW and Politico sued Trump and other members of his administration, saying the ban violated the US Constitution’s First Amendment protections for free speech and a free press, as well as their due process rights. A coalition of press freedom groups and dozens of news organizations, including Reuters, later filed a legal brief supporting the action.

Despite the temporary ruling restoring access, the court has not yet resolved the main challenge to the ban, based on the First Amendment of the US Constitution that protects the right to free speech and a free press. The news outlets now have until September 28 to ask the judge to keep their access in place while the case moves forward.

Even if the outlets won a permanent injunction against the ban, Trump could still pursue his case in federal court.

Trump has waged an unprecedented campaign against the media since returning to office last year, launching personal attacks on journalists, denying access to some outlets and suing news organisations, as well as calling on the Federal Communications Commission to strip broadcast stations of their licenses.

Friendliest press corps

On Thursday, Trump made a joke to photographers in the Oval Office about China — which imposes strict censorship across its media — having the friendliest press corps, as he welcomed Xi to the White House.

While the White House on Thursday told the court it had reinstated CNN, MS NOW and Politico’s access, CNN and MS NOW later said some of their journalists were denied access to cover Xi’s arrival for the White House state dinner, according to the Washington Post.

A lawyer for the three outlets in court this week warned that the White House could again try to revoke the outlets’ White House access as the current litigation continues.

David Ellison, CEO of Paramount Skydance, which is seeking to merge with CNN-parent Warner Bros. Discovery, attended the Trump-Xi state dinner Thursday night.

Critics have raised concerns about Paramount-owned CBS News tailoring coverage to favor Trump, although Ellison has promised to create an editorial independence board for both CBS and CNN as part of the deal.

The latest example came on Friday when OpenAI said its agents had leaked 53 images from ChatGPT users. OpenAI declined to say if the images were AI-generated or identified real people. It also declined to say when the images were posted.

The disclosure and researcher reports on Friday of other previously unknown activity involving several US agencies reveal a new area of privacy risk for the company, and illustrate how difficult it is even for an AI firm at the cutting edge of the technology to inventory all the unauthorized activity tied to its agents.

OpenAI’s ongoing battle also reflects a yawning gap between the strength of the models the company is testing and its capacity to oversee or even track their actions.

As of mid-September, one person briefed on the matter estimated that OpenAI had found roughly two dozen incidents of its agents acting in undesirable ways. But the number has continued rising as OpenAI teams sift through internal logs of the agents’ activities and find previously unknown cases, the two people close to the company said.

OpenAI said its review would take months to complete given the scale of the work.

The company also said it had notified dozens of third parties about improper activity.

Most of the leaked images have been taken down and OpenAI said it was lobbying hosting providers to remove the rest.

OpenAI’s agents had access to these images because the company relies on anonymised user data for part of its model-training process, according to the company, former employees and outside researchers. Enterprise data is not eligible for training, while ChatGPT consumers need to opt out of allowing the company to use their data for training.

Before user posts are used for training, they go through an anonymisation process that strips out metadata, names and other contact information and should make it difficult to trace back to any individual user, the company said.

But the practice carries risks because there is a chance that the data may not be fully stripped of personally identifiable information and that it might leak in the course of the models work, three people familiar with OpenAI’s practices said.

OpenAI agents accessed US websites

OpenAI said late on Friday its models accessed information from the websites of the US Securities and Exchange Commission and the US Census Bureau during research and training activity, but found no evidence of unauthorized access, compromised accounts or security breaches.

Separately, AI research nonprofit Transluce said agents appearing to originate from OpenAI made an unsuccessful attempt to hack a US Department of Education civil rights website.

Transluce said the incident was part of broader AI agent activity probing government websites using tactics including exposed credentials, anti-bot bypasses and fake accounts.

More than 15 cases

In the two months since OpenAI first announced that its agents broke containment, there have been more than 15 different OpenAI-related incidents of varying levels of severity disclosed by the company, by outside researchers, or just on Wednesday by Australian Prime Minister Anthony Albanese at the United Nations, who said OpenAI agents broke into a government health data portal in June.

Past incidents have varied in nature, ranging from spam-like messages left on internet sites all the way to the break-in at Hugging Face, which involved a swarm of agents abusing previously unknown software vulnerabilities to escape their networks and penetrate the AI repository as they hunted for answers to a test. OpenAI also said its agents took aim at its own infrastructure.

Albanese told reporters in New York that OpenAI uncovered the activity in August, and disclosed it on September 10 via an email to a general government inbox. He said he directly told OpenAI CEO Sam Altman that this disclosure process was unacceptable.

OpenAI said some of the sites involved are operated by government, universities and public agencies because the models that are conducting research seek out reputable sources of public information.

A locked-down process

The July 21 announcement that OpenAI’s agents had slipped out of control and hacked Hugging Face sparked widespread worries within the AI industry over its ability to control the more powerful AI models under development now.

Since then, Anthropic, Alphabet’s Google and Meta have said they’ve found similar behavior by their agents after the Hugging Face incident prompted them to search.

OpenAI has acknowledged a general need for more transparency around rogue AI behavior. On September 16, the company published a new framework for disclosing such incidents, saying it would err on the side of transparency even when significance is uncertain.

Even so, two people familiar with OpenAI’s investigation into its agents activity described it as locked down and shaped by company lawyers. The process has been unusually compartmentalized for a company that some former employees say was more open about these issues in the past, the people said.

Roughly 100 people were in some way involved in the process to understand the Hugging Face hack, three people briefed on the matter said. During that process, evidence of other incidents surfaced.

Reuters has previously reported that OpenAI investigators looking into the Hugging Face breach were discouraged by the company’s lawyers from expanding the scope of the investigation to include other incidents. OpenAI said its lawyers did not discourage deeper investigation.

Many incidents have been uncovered by outside researchers rather than OpenAI directly. In several episodes, the agents took problematic actions that went unnoticed by the company for months.

Earlier this month, a small group of investigators discovered that the company’s agents had hijacked a mostly defunct German wiki site to share tactics to cheat on some tasks, bypass OpenAIs restrictions and mask their behavior.

This week, the AI research firm Transluce said it discovered that OpenAI agents had bypassed the Australian Institute of Health and Welfare’s anti-bot controls. The firm also found two other cases that it linked to OpenAI agents. Those incidents were separate from the activity disclosed by Albanese.

In a statement, OpenAI said much of the activity described in Transluces report overlaps with cases at varying stages of investigation in our ongoing review of misaligned model activity. The company said it is prioritising the most severe cases in its review.

Since the Hugging Face hack, researchers across the AI industry have grown worried that companies will not be able to predict or control their technology. Some have taken the path of former Anthropic researcher Jacob Coxon, who publicly resigned this month in a viral social-media thread that said the AI labs are gambling with our lives.

In response to those concerns, Altman and his counterpart at Anthropic, CEO Dario Amodei, called for the industry to pace the development of AI and move cautiously in its pursuit of recursive self improvement. Altman doubled down on that message this week while addressing the UN.

Even so, both companies rolled out new models on Tuesday.

The ordinance aims to protect Safe City cameras and surveillance systems from damage and ensure compensation for any damage caused to them.

Under the ordinance, financial penalties may be imposed for damaging Safe City cameras and surveillance infrastructure, as well as for inciting or encouraging such damage.

The President’s Office announced the approval in a post on X, saying the ordinance was titled the Islamabad Capital Territory Safe City Camera and Surveillance Infrastructure Protection and Damage Compensation Ordinance.

In case of damage, a fine of up to Rs2 million may be imposed for each camera or piece of infrastructure. If the damage occurs during a public meeting, procession, sit-in, protest, strike or gathering, the organisers may also be held responsible, the Presidency added.

The amount of the fine and the cost of the damage will be determined by an authorised officer after providing the person concerned with an opportunity to be heard.

It further said the person concerned would have the right to appeal against the determination of the fine or cost of the damage before the District and Sessions Judge, Islamabad, within 30 days.

In case of non-payment of the prescribed fine or compensation, the amount may be recovered as arrears of land revenue. The amount recovered in the form of fines and compensation will be spent on the repair and replacement of Safe City cameras and surveillance infrastructure, it added.

The move to protect the infrastructure of the capital comes after the PTI deferred its planned protest to Islamabad until Oct 4 for the release of its founder, Imran Khan, who is currently incarcerated in Rawalpindi’s Adiala jail.

Despite a large contingent of law enforcement personnel and the party’s claims of a peaceful protest, the miscreants damaged 441 cameras of Safe City Authority during the march to Islamabad in 2024, according to the officials.

The commitment was made at a meeting of Chief of Defence Forces (CDF) and Chief of Army Staff (COAS) Field Marshal Asim Munir and the chiefs of general staff of Saudi Arabia and Turkiye in Riyadh on Friday.

The meeting took place against the backdrop of heightened security concerns in the kingdom, with the Houthis and the Yemeni government engaged in renewed fighting. The Houthis have declared a maritime blockade targeting Saudi Arabia and attacked ships in the Red Sea, while also carrying out attacks in Saudi Arabia in recent weeks.

Earlier, the Foreign Office (FO) had said the three countries were considering a meeting of their military chiefs to discuss ways to assist Riyadh under the Makkah Defence Alliance.

According to the ISPR, the participants discussed measures to advance cooperation among their armed forces, including enhanced intelligence sharing, military coordination and integration of capabilities.

The measures would help strengthen deterrence and collective defence, the military’s media wing said.

The participants also reviewed the security situation and threats facing Saudi Arabia and strongly “resented” attacks targeting the holy city of Makkah as well as civilian and economic infrastructure.

They appreciated the vigilance and professionalism of the Saudi armed forces in ensuring the safety of the people and security of the kingdom, according to the ISPR.

The military chiefs further affirmed their commitment to collective defence, which, the ISPR said, demonstrated the resolve of the three countries to defend Saudi Arabia’s security, sovereignty and territorial integrity.

The ISPR said the Riyadh meeting reflected the strength of the defence partnership among Pakistan, Saudi Arabia and Turkiye, as well as their shared determination to implement the commitments under the tripartite defence agreement.

The participants also commended the longstanding relations and mutual trust between the three countries and their armed forces.

The Makkah alliance was signed by Pakistan, Saudi Arabia and Turkiye in Makkah on Aug 7 and the first meeting of the Strategic Political and Defence Committee was subsequently held in Istanbul on Aug 31, where the three countries agreed to establish a permanent secretariat in Riyadh and decided that Pakistan would provide the secretary general for its first three-year term.

Pakistan appointed retired Lt Gen Nauman Mahmood as the alliance’s first secretary general, taking another step towards putting the institutional structure of the pact in place.