News of the Cyber World

[kyonides]

OpenAI, Google, Microsoft Join Biden Administration ‘AI Safety’ Consortium

Mashable reports that the Biden administration has officially tapped major tech players and other AI stakeholders to address safety and trust in AI development. On Thursday, the U.S. Department of Commerce announced the creation of the AI Safety Institute Consortium (AISIC).

The consortium, housed under the Commerce Department’s National Institute of Standards and Technology (NIST), will follow President Biden’s AI executive order mandates. This includes “developing guidelines for red-teaming, capability evaluations, risk management, safety and security, and watermarking synthetic content,” said Secretary of Commerce Gina Raimondo.

Over 200 participants have joined AISIC so far. This includes AI leaders like ChatGPT developer OpenAI, Google, Microsoft, Apple, Amazon, Mark Zuckerberg’s Meta, and NVIDIA. Academics from MIT, Stanford, Cornell, and others are also participating. Additionally, industry researchers and think tanks like the Center for AI Safety, IEEE, and the Responsible AI Institute are involved.

The consortium stems from Biden’s sweeping executive order seeking to regulate AI development. “The government has a role in setting standards and tools to mitigate AI risks and harness its potential,” said Raimondo.

While the EU has worked on AI regulations, this marks a major U.S. government effort to formally oversee AI. “I understand this is a first step by the administration to work with industry on some of the hard problems. But it is an important first step and I think this is an area where industry and government collaboration is critical,” said Microsoft President Brad Smith.

Meta Bans Iran’s Supreme Leader Ali Khamenei from Facebook, Instagram for Call to ‘Wipe Israel Off the Map’

The New York Post reports that Meta, the parent company of Facebook and Instagram, has permanently banned Iranian Supreme Leader Ali Khamenei from its platforms. The company stated this action was taken in response to Khamenei’s repeated calls to “wipe Israel off the map” and referring to Israel as a “cancerous tumor” that will be “uprooted and destroyed.”

Khamenei’s Persian and English-language Instagram accounts, which had over 5 million followers combined, along with his official Facebook page, were removed for violating Meta’s rules against dangerous individuals and organizations. Meta said Khamenei’s comments qualified as hate speech that justified the ban.

The Iranian despot has long used inflammatory language against Israel, which Iran considers its top adversary. In October 2022, following attacks between Israel and Hamas that left over 1,000 dead, Khamenei wrote on Twitter that “the cancer of the usurper Zionist regime will be eradicated at the hands of the Palestinian people and the Resistance forces throughout the region.”

While Twitter has not banned Khamenei, pro-Israel groups have been lobbying social media platforms to take down his accounts due to repeated calls for violence. Facebook, Instagram, and Twitter are all blocked within Iran, but Khamenei maintained accounts on these platforms to reach international audiences.

Mark Zuckerberg’s Facebook, Instagram to Start Flagging AI-Generated Images as Elections Ramp Up

The Verge reports that the rise of AI-generated media has prompted Zuckerberg to roll out new labeling for AI content across his apps. In a blog post, Meta announced that in an effort to get ahead of potential misuse, the company will add watermarks to AI imagery on Facebook, Instagram and Threads. For more deceptive uses, Meta plans to penalize users who do not disclose realistic AI video or audio.

Meta President Nick Clegg announced the moves Tuesday as election seasons ramp up globally. “For those who are worried about video, audio content being designed to materially deceive the public on a matter of political importance in the run-up to the election, we’re going to be pretty vigilant,” Clegg said. “Do I think that there is a possibility that something may happen where, however quickly it’s detected or quickly labeled, nonetheless we’re somehow accused of having dropped the ball? Yeah, I think that is possible, if not likely.”

Meta is joining the companies by implementing content protection measures similar to Adobe’s Content Credentials metadata system. Google also recently expanded its SynthID watermark feature to include audio. Meta will soon require users to transparently note realistic AI video or audio. Violators face “warnings through to removal,” per Clegg. Highlighting examples of viral AI-generated political posts, Clegg believes widespread deception is unlikely. “I just don’t think that’s the way that it’s going to play out,” he said.

Internally, Meta has started testing large language models trained on its content policies to effectively identify violations. Clegg called it an efficient triage mechanism for human moderators. “It appears to be a highly effective and rather precise way of ensuring that what is escalated to our human reviewers really is the kind of edge cases for which you want human judgment,” Clegg said.

Driverless Car from Google’s Waymo Strikes Cyclist in San Francisco

The Verge reports that a self-driving car built by Google’s Waymo struck and injured a cyclist in San Francisco on the afternoon of Tuesday February 6. The incident occurred at the intersection of 17th and Mississippi Streets in the Potrero Hill neighborhood. According to the San Francisco Police Department (SFPD), the autonomous vehicle collided with a person on a bicycle, causing non-life threatening injuries.

The Waymo vehicle was stopped at a four-way stop when a large truck began turning into the intersection, according to a statement from Waymo spokesperson Julia Ilina. The Waymo car waited for its turn and then proceeded into the intersection, failing to identify the cyclist who was traveling behind the truck.

“The cyclist was occluded by the truck and quickly followed behind it, crossing into the Waymo vehicle’s path,” Ilina explained. “When they became fully visible, our vehicle applied heavy braking but was not able to avoid the collision.”

Waymo contacted the police, but the injured cyclist left the scene on their own, reporting only minor scratches. The passenger inside the Waymo vehicle was unharmed. The SFPD says the incident remains under investigation.