×
The Final Season of ‘Stranger Things’ Teases Its Lengthy Runtimes

The Final Season of ‘Stranger Things’ Teases Its Lengthy Runtimes

Tech-news , , , ,

Marvel TV’s big boss teases a Foggy appearance in Daredevil: Born Again season 2. Guillermo del Toro may bring The Buried Giant to stop-motion life at Netflix. Plus, get a few more looks at Predator Badlands. To me, my spoilers!

Five Nights at Freddy’s 2

Variety reports Megan Fox has joined Five Nights at Freddy’s 2 as the voice of Toy Chica. Additionally, MatPat will voice Toy Bonnie and Kellen Goff will voice Toy Freddy.

The Buried Giant

Bloody-Disgusting also has word that a stop-motion animated film based on Kazuo Ishiguro’s fantasy novel, The Buried Giant, is now in development at Netflix with Guillermo del Toro attached to direct. The story follows follows “an elderly Briton couple, Axl and Beatrice, living in a fictional post-Arthurian England in which no one is able to retain long-term memories.”

Predator: Badlands

Fangoria has four new posters for Predator: Badlands.

G3jkapjauaagal0.jpg
© 20th Century Studios
G3jovaqaeaaj 12.jpg
© 20th Century Studios
G3jrupdaqaa0xlp.jpg
© 20th Century Studios
G3jyvhjbyaau2q6.jpg
© 20th Century Studios

Deathstalker

Composers Chuck Cirino, Bear McCreary and Slash have released a music video for the theme song to Steven Kostankski’s Deathstalker remake with singer Brendan McCreary, drummer Gene Hoglan, bassist Pete Griffin, and guitarist Omer Ben-Zvi.

Yellowjackets

In a new press release (via TV Line), Yellowjackets creators Ashley Lyle and Bart Nickerson revealed the series will conclude with its fourth season.

After three incredible seasons, and great consideration, we’re excited to announce that we will be bringing the story of Yellowjackets to its twisted conclusion in this fourth and final season. We’ve always known there would come a point when the story would tell us it wants to end, and it’s our belief that our job — our responsibility — is to listen. Telling this emotional, wild, and deeply human story has been a profoundly meaningful experience and a true honor for us, and we’re so very grateful to the brilliant cast, crew and writers who have bravely gone on the journey with us to bring it to life. Most of all, we want to thank the fans who have stuck with us through every moment, mystery and meal — the Hive is nothing without you! We can’t wait to share the final chapter with you and hope you find it…delicious.

The Creep Tapes

However, The Creep Tapes has been renewed for a third season at Shudder. [Fangoria]

Magic 8 Ball

Deadline also has word M. Night Shyamalan will direct a TV series for Mattel based on its “Magic 8 Ball” fortune-telling toy. Co-created with Bad Falcuk, the series aims to “reimagine the classic Magic 8 Ball as the centerpiece of a high-concept, character-driven supernatural drama that blends psychological intensity with cultural intrigue.”

Daredevil: Born Again

During a recent interview with Entertainment Weekly, Daredevil: Born Again producer Brad Winderbaum confirmed Elden Henson does appear as Foggy Nelson in the second season of Daredevil: Born Again.

I actually haven’t seen the quote, but I can tell you he is in season 2. People don’t know what they’re allowed to say and what they’re not allowed to say, also. Do I play it coy? Do I be honest? It’s always a dance.

In the same interview, Winderbaum also revealed Matthew Lillard plays a character named Mr. Charles.

He plays a character named Mr. Charles. The story is about power. When Wilson Fisk takes over New York, not just as a mayor but as a king in a way, it puts him in a new class of power players on the international stage. Matthew Lillard’s character represents that. So he’s a bit of a new antagonist on the field, but he’s as influential in many ways as Fisk is. That power negotiation at a very high level of politics and international diplomacy is also fun to watch. He is dealing with Mr. Charles up here and Daredevil down here for different reasons. So he’s getting squeezed a little bit.

The Walking Dead: Dead City

According to Deadline, Raúl Castillo has joined the cast of The Walking Dead: Dead City as a character named Luis.

Prism

Deadline reports Millie Bobby Brown will star in Prism at Netflix, a new supernatural series produced by Rachel Brosnahan, Etan Frankel, Anthony Russo and Joe Russo, and Brown herself. Based on a short story by Nick Shafir, the series sees Brown as Cassie, “a woman with the unique ability to communicate with apparitions, who must uncover the cause of a newly discovered phenomenon that causes ‘visitors’ (ghosts) to appear all over the world before it’s too late.”

Stranger Things

According to Ross Duffer on Instagram, the fifth season premiere of Stranger Things is titled “The Crawl” and runs 68 minutes long. Episode 2 is titled “The Vanishing“ and runs 54 minutes, while Episode 3, “The Turnbow Trap”, is 66 minutes. Episode 4, “Sorcerer”, runs 83 minutes.

The Copenhagen Test

Finally, Simu Liu learns “his brain has been hacked”  by a “shadowy agency” in the trailer for The Copenhagen Test.

 

 

Want more io9 news? Check out when to expect the latest Marvel, Star Wars, and Star Trek releases, what’s next for the DC Universe on film and TV, and everything you need to know about the future of Doctor Who.

Source link
#Final #Season #Stranger #Teases #Lengthy #Runtimes

Tag

Related Posts

X is overhauling how it pays creators, phasing out its long-running Revenue Sharing program in favor of a new system called the Original Content Rewards Program.

The platform’s Creators account announced the change in a post, saying the program is meant to “reward creators who bring original ideas, expertise, reporting, creativity, and commentary to X.”

According to the announcement, X stopped accepting new Revenue Sharing enrollments immediately and will wind the program down entirely after Sept. 7. Creators currently enrolled will receive three final payouts — two on the standard schedule in mid- and late August, and a final payment around Sept. 11 covering earnings through the cutoff date. Starting Sept. 8, existing Revenue Sharing members can apply for the new program if they meet its eligibility requirements, the company said.

SEE ALSO: With the launch of X Money, Elon Musk gets one step closer to his ‘everything app’

Per X’s announcement, the new program ties earnings to “qualified impressions” a creator’s original content receives, rather than the advertising-revenue split that defined the old system. Qualified impressions must come from unique, verified Premium subscribers viewing at least half of a post on the Home Timeline; impressions that are duplicated, paid, promoted, or fraudulent don’t count.

To qualify, X says creators must be at least 18, live in an eligible country, hold a Personal or Business account in good standing, subscribe to a paid X tier, and have at least 500 verified followers, along with 500,000 Home Timeline impressions from verified users over the prior 90 days. Those requirements must be maintained continuously to keep receiving payouts, according to the announcement.

X’s guidelines lean heavily on defining what counts as “original.” Per the company’s post, content that is copied, reuploaded without authorship, generated through automated means, or reposted with only minor edits — such as captions, watermarks, or basic text overlays — won’t qualify. Users can earn from reposting others’ material only if they add substantive commentary, analysis, or creative editing. So, simply describing what’s already happening in a post doesn’t meet the bar.

As noted in Engadget’s report on the news, X adjusted its Revenue Sharing formula in March to weight engagement more heavily toward a creator’s home region. That shift, they wrote, likely followed revelations that a number of popular accounts posting pro-Trump content and US-focused commentary were not actually based in the United States.

Whether the new Original Content Rewards Program carries over a similar regional weighting is not yet clear.

#retires #revenue #sharing #Original #Content #Rewards #program">X retires revenue sharing for new ‘Original Content Rewards’ program X is overhauling how it pays creators, phasing out its long-running Revenue Sharing program in favor of a new system called the Original Content Rewards Program. The platform’s Creators account announced the change in a post, saying the program is meant to “reward creators who bring original ideas, expertise, reporting, creativity, and commentary to X.”According to the announcement, X stopped accepting new Revenue Sharing enrollments immediately and will wind the program down entirely after Sept. 7. Creators currently enrolled will receive three final payouts — two on the standard schedule in mid- and late August, and a final payment around Sept. 11 covering earnings through the cutoff date. Starting Sept. 8, existing Revenue Sharing members can apply for the new program if they meet its eligibility requirements, the company said. SEE ALSO: With the launch of X Money, Elon Musk gets one step closer to his ‘everything app’ Per X’s announcement, the new program ties earnings to “qualified impressions” a creator’s original content receives, rather than the advertising-revenue split that defined the old system. Qualified impressions must come from unique, verified Premium subscribers viewing at least half of a post on the Home Timeline; impressions that are duplicated, paid, promoted, or fraudulent don’t count. Mashable Light Speed To qualify, X says creators must be at least 18, live in an eligible country, hold a Personal or Business account in good standing, subscribe to a paid X tier, and have at least 500 verified followers, along with 500,000 Home Timeline impressions from verified users over the prior 90 days. Those requirements must be maintained continuously to keep receiving payouts, according to the announcement.X’s guidelines lean heavily on defining what counts as “original.” Per the company’s post, content that is copied, reuploaded without authorship, generated through automated means, or reposted with only minor edits — such as captions, watermarks, or basic text overlays — won’t qualify. Users can earn from reposting others’ material only if they add substantive commentary, analysis, or creative editing. So, simply describing what’s already happening in a post doesn’t meet the bar. As noted in Engadget’s report on the news, X adjusted its Revenue Sharing formula in March to weight engagement more heavily toward a creator’s home region. That shift, they wrote, likely followed revelations that a number of popular accounts posting pro-Trump content and US-focused commentary were not actually based in the United States. Whether the new Original Content Rewards Program carries over a similar regional weighting is not yet clear. #retires #revenue #sharing #Original #Content #Rewards #program

announced the change in a post, saying the program is meant to “reward creators who bring original ideas, expertise, reporting, creativity, and commentary to X.”

According to the announcement, X stopped accepting new Revenue Sharing enrollments immediately and will wind the program down entirely after Sept. 7. Creators currently enrolled will receive three final payouts — two on the standard schedule in mid- and late August, and a final payment around Sept. 11 covering earnings through the cutoff date. Starting Sept. 8, existing Revenue Sharing members can apply for the new program if they meet its eligibility requirements, the company said.

SEE ALSO: With the launch of X Money, Elon Musk gets one step closer to his ‘everything app’

Per X’s announcement, the new program ties earnings to “qualified impressions” a creator’s original content receives, rather than the advertising-revenue split that defined the old system. Qualified impressions must come from unique, verified Premium subscribers viewing at least half of a post on the Home Timeline; impressions that are duplicated, paid, promoted, or fraudulent don’t count.

To qualify, X says creators must be at least 18, live in an eligible country, hold a Personal or Business account in good standing, subscribe to a paid X tier, and have at least 500 verified followers, along with 500,000 Home Timeline impressions from verified users over the prior 90 days. Those requirements must be maintained continuously to keep receiving payouts, according to the announcement.

X’s guidelines lean heavily on defining what counts as “original.” Per the company’s post, content that is copied, reuploaded without authorship, generated through automated means, or reposted with only minor edits — such as captions, watermarks, or basic text overlays — won’t qualify. Users can earn from reposting others’ material only if they add substantive commentary, analysis, or creative editing. So, simply describing what’s already happening in a post doesn’t meet the bar.

As noted in Engadget’s report on the news, X adjusted its Revenue Sharing formula in March to weight engagement more heavily toward a creator’s home region. That shift, they wrote, likely followed revelations that a number of popular accounts posting pro-Trump content and US-focused commentary were not actually based in the United States.

Whether the new Original Content Rewards Program carries over a similar regional weighting is not yet clear.

#retires #revenue #sharing #Original #Content #Rewards #program">X retires revenue sharing for new ‘Original Content Rewards’ program

X is overhauling how it pays creators, phasing out its long-running Revenue Sharing program in favor of a new system called the Original Content Rewards Program.

The platform’s Creators account announced the change in a post, saying the program is meant to “reward creators who bring original ideas, expertise, reporting, creativity, and commentary to X.”

According to the announcement, X stopped accepting new Revenue Sharing enrollments immediately and will wind the program down entirely after Sept. 7. Creators currently enrolled will receive three final payouts — two on the standard schedule in mid- and late August, and a final payment around Sept. 11 covering earnings through the cutoff date. Starting Sept. 8, existing Revenue Sharing members can apply for the new program if they meet its eligibility requirements, the company said.

SEE ALSO: With the launch of X Money, Elon Musk gets one step closer to his ‘everything app’

Per X’s announcement, the new program ties earnings to “qualified impressions” a creator’s original content receives, rather than the advertising-revenue split that defined the old system. Qualified impressions must come from unique, verified Premium subscribers viewing at least half of a post on the Home Timeline; impressions that are duplicated, paid, promoted, or fraudulent don’t count.

To qualify, X says creators must be at least 18, live in an eligible country, hold a Personal or Business account in good standing, subscribe to a paid X tier, and have at least 500 verified followers, along with 500,000 Home Timeline impressions from verified users over the prior 90 days. Those requirements must be maintained continuously to keep receiving payouts, according to the announcement.

X’s guidelines lean heavily on defining what counts as “original.” Per the company’s post, content that is copied, reuploaded without authorship, generated through automated means, or reposted with only minor edits — such as captions, watermarks, or basic text overlays — won’t qualify. Users can earn from reposting others’ material only if they add substantive commentary, analysis, or creative editing. So, simply describing what’s already happening in a post doesn’t meet the bar.

As noted in Engadget’s report on the news, X adjusted its Revenue Sharing formula in March to weight engagement more heavily toward a creator’s home region. That shift, they wrote, likely followed revelations that a number of popular accounts posting pro-Trump content and US-focused commentary were not actually based in the United States.

Whether the new Original Content Rewards Program carries over a similar regional weighting is not yet clear.

#retires #revenue #sharing #Original #Content #Rewards #program
Buc-ee’s became something of a viral sensation during the World Cup, but it has a troubling history of suing small gas stations and convenience stores. On a recent episode of Last Week Tonight, John Oliver literally begged the company to sue him for selling merch featuring his squirrel mascot, Mr. Nutterbutter, with branding that reads “Buc-Off.” But Buc-ee’s has so far declined to take the bait, instead targeting yet another small business that lacks the deep pockets and legal team of HBO.

According to WDTN, an NBC affiliate in Miami Valley, Ohio, Buc-ee’s opened its first locations in the state earlier this year. And now it’s targeting established local businesses through trademark suits. Beaver’s Mini Mart in Beavercreek has been a staple of the city for “decades.” But Buc-ee’s is claiming that this long-running store’s beaver logo could confuse customers. WDTN says:

In the suit, filed days ago, Buc-ee’s alleges that the Mini Mart’s cartoon beaver mascot is too similar to their own, with its “wide eyes and a smile” that also “uses red as a predominant color,” and could cause confusion.

Oliver and Mr. Nutterbutter have already fended off one lawsuit, with Bob Murray, a mining executive, having unsuccessfully sued for defamation following a 2017 episode about the coal industry.

#Bucees #dodges #John #Oliver #sue #small #businessBusiness,Culture,Entertainment,Internet Culture,Law,News,Policy,TV Shows">Buc-ee’s dodges John Oliver to sue another small businessBuc-ee’s became something of a viral sensation during the World Cup, but it has a troubling history of suing small gas stations and convenience stores. On a recent episode of Last Week Tonight, John Oliver literally begged the company to sue him for selling merch featuring his squirrel mascot, Mr. Nutterbutter, with branding that reads “Buc-Off.” But Buc-ee’s has so far declined to take the bait, instead targeting yet another small business that lacks the deep pockets and legal team of HBO.According to WDTN, an NBC affiliate in Miami Valley, Ohio, Buc-ee’s opened its first locations in the state earlier this year. And now it’s targeting established local businesses through trademark suits. Beaver’s Mini Mart in Beavercreek has been a staple of the city for “decades.” But Buc-ee’s is claiming that this long-running store’s beaver logo could confuse customers. WDTN says:In the suit, filed days ago, Buc-ee’s alleges that the Mini Mart’s cartoon beaver mascot is too similar to their own, with its “wide eyes and a smile” that also “uses red as a predominant color,” and could cause confusion.Oliver and Mr. Nutterbutter have already fended off one lawsuit, with Bob Murray, a mining executive, having unsuccessfully sued for defamation following a 2017 episode about the coal industry.#Bucees #dodges #John #Oliver #sue #small #businessBusiness,Culture,Entertainment,Internet Culture,Law,News,Policy,TV Shows

viral sensation during the World Cup, but it has a troubling history of suing small gas stations and convenience stores. On a recent episode of Last Week Tonight, John Oliver literally begged the company to sue him for selling merch featuring his squirrel mascot, Mr. Nutterbutter, with branding that reads “Buc-Off.” But Buc-ee’s has so far declined to take the bait, instead targeting yet another small business that lacks the deep pockets and legal team of HBO.

According to WDTN, an NBC affiliate in Miami Valley, Ohio, Buc-ee’s opened its first locations in the state earlier this year. And now it’s targeting established local businesses through trademark suits. Beaver’s Mini Mart in Beavercreek has been a staple of the city for “decades.” But Buc-ee’s is claiming that this long-running store’s beaver logo could confuse customers. WDTN says:

In the suit, filed days ago, Buc-ee’s alleges that the Mini Mart’s cartoon beaver mascot is too similar to their own, with its “wide eyes and a smile” that also “uses red as a predominant color,” and could cause confusion.

Oliver and Mr. Nutterbutter have already fended off one lawsuit, with Bob Murray, a mining executive, having unsuccessfully sued for defamation following a 2017 episode about the coal industry.

#Bucees #dodges #John #Oliver #sue #small #businessBusiness,Culture,Entertainment,Internet Culture,Law,News,Policy,TV Shows">Buc-ee’s dodges John Oliver to sue another small business

Buc-ee’s became something of a viral sensation during the World Cup, but it has a troubling history of suing small gas stations and convenience stores. On a recent episode of Last Week Tonight, John Oliver literally begged the company to sue him for selling merch featuring his squirrel mascot, Mr. Nutterbutter, with branding that reads “Buc-Off.” But Buc-ee’s has so far declined to take the bait, instead targeting yet another small business that lacks the deep pockets and legal team of HBO.

According to WDTN, an NBC affiliate in Miami Valley, Ohio, Buc-ee’s opened its first locations in the state earlier this year. And now it’s targeting established local businesses through trademark suits. Beaver’s Mini Mart in Beavercreek has been a staple of the city for “decades.” But Buc-ee’s is claiming that this long-running store’s beaver logo could confuse customers. WDTN says:

In the suit, filed days ago, Buc-ee’s alleges that the Mini Mart’s cartoon beaver mascot is too similar to their own, with its “wide eyes and a smile” that also “uses red as a predominant color,” and could cause confusion.

Oliver and Mr. Nutterbutter have already fended off one lawsuit, with Bob Murray, a mining executive, having unsuccessfully sued for defamation following a 2017 episode about the coal industry.

#Bucees #dodges #John #Oliver #sue #small #businessBusiness,Culture,Entertainment,Internet Culture,Law,News,Policy,TV Shows
OpenAI said Friday it has suspended work on some aspects of its upcoming model Astra after an internal review found it had made significant advancements in agentic coding and cybersecurity — enough to warrant concern over its capabilities.

OpenAI said in a blog post Friday that this model, which is still in development, reached its “critical cybersecurity threshold,” meaning it could independently identify and carry out cyberattacks against traditionally well-protected real-world systems. Under the company’s “Preparedness Framework,” which it created in 2023, this triggered additional safeguards.

“While we continue to benchmark and assess this model, our preliminary evaluations indicate strong enough performance that we cannot rule out Critical capability level at this time,” OpenAI wrote. “Astra is an upcoming model, and was not involved in exploiting Hugging Face.”

The disclosure highlights an unusual moment in the topsy-turvy and still nascent frontier AI labs sector. Companies across every industry hold back products over potential risks, including for safety and cybersecurity concerns. But they rarely announce those decisions publicly when it’s a product that is still under development.

In this case, OpenAI is already under scrutiny after a different unreleased model breached Hugging Face’s systems during internal testing — the first verifiable incident of an AI lab losing control of its model. Since then, OpenAI and AI labs such as Anthropic have disclosed other incidents in which AI models breached their sandboxes and posed threats during cybersecurity tests.

The string of cases — seems like a new disclosure every day now — has triggered varying reactions from cybersecurity experts, lawmakers, and the AI labs themselves. Some express fear and call for stricter oversight. But there’s also a bit of flexing. In certain circles, any AI lab with a model that has that kind of capability will be seen as an impressive advancement.

OpenAI said it was sharing this information because it believes “it’s important to be transparent with the public and the safety and security communities about this potential shift in capabilities.”

The AI lab said it’s also taking action, including enacting stricter security controls and pausing internal activities involving Astra that don’t meet these beefed guardrails. OpenAI said it is working with relevant government agencies and “select AI safety organizations” to test the capabilities for this model.

When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.

#OpenAI #slowed #Astra #model #development #security #concerns #TechCrunchOpenAI">OpenAI says it slowed Astra model development over security concerns | TechCrunch OpenAI said Friday it has suspended work on some aspects of its upcoming model Astra after an internal review found it had made significant advancements in agentic coding and cybersecurity — enough to warrant concern over its capabilities. OpenAI said in a blog post Friday that this model, which is still in development, reached its “critical cybersecurity threshold,” meaning it could independently identify and carry out cyberattacks against traditionally well-protected real-world systems. Under the company’s “Preparedness Framework,” which it created in 2023, this triggered additional safeguards. “While we continue to benchmark and assess this model, our preliminary evaluations indicate strong enough performance that we cannot rule out Critical capability level at this time,” OpenAI wrote. “Astra is an upcoming model, and was not involved in exploiting Hugging Face.” The disclosure highlights an unusual moment in the topsy-turvy and still nascent frontier AI labs sector. Companies across every industry hold back products over potential risks, including for safety and cybersecurity concerns. But they rarely announce those decisions publicly when it’s a product that is still under development. In this case, OpenAI is already under scrutiny after a different unreleased model breached Hugging Face’s systems during internal testing — the first verifiable incident of an AI lab losing control of its model. Since then, OpenAI and AI labs such as Anthropic have disclosed other incidents in which AI models breached their sandboxes and posed threats during cybersecurity tests. The string of cases — seems like a new disclosure every day now — has triggered varying reactions from cybersecurity experts, lawmakers, and the AI labs themselves. Some express fear and call for stricter oversight. But there’s also a bit of flexing. In certain circles, any AI lab with a model that has that kind of capability will be seen as an impressive advancement. OpenAI said it was sharing this information because it believes “it’s important to be transparent with the public and the safety and security communities about this potential shift in capabilities.” The AI lab said it’s also taking action, including enacting stricter security controls and pausing internal activities involving Astra that don’t meet these beefed guardrails. OpenAI said it is working with relevant government agencies and “select AI safety organizations” to test the capabilities for this model. When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.#OpenAI #slowed #Astra #model #development #security #concerns #TechCrunchOpenAI

blog post Friday that this model, which is still in development, reached its “critical cybersecurity threshold,” meaning it could independently identify and carry out cyberattacks against traditionally well-protected real-world systems. Under the company’s “Preparedness Framework,” which it created in 2023, this triggered additional safeguards.

“While we continue to benchmark and assess this model, our preliminary evaluations indicate strong enough performance that we cannot rule out Critical capability level at this time,” OpenAI wrote. “Astra is an upcoming model, and was not involved in exploiting Hugging Face.”

The disclosure highlights an unusual moment in the topsy-turvy and still nascent frontier AI labs sector. Companies across every industry hold back products over potential risks, including for safety and cybersecurity concerns. But they rarely announce those decisions publicly when it’s a product that is still under development.

In this case, OpenAI is already under scrutiny after a different unreleased model breached Hugging Face’s systems during internal testing — the first verifiable incident of an AI lab losing control of its model. Since then, OpenAI and AI labs such as Anthropic have disclosed other incidents in which AI models breached their sandboxes and posed threats during cybersecurity tests.

The string of cases — seems like a new disclosure every day now — has triggered varying reactions from cybersecurity experts, lawmakers, and the AI labs themselves. Some express fear and call for stricter oversight. But there’s also a bit of flexing. In certain circles, any AI lab with a model that has that kind of capability will be seen as an impressive advancement.

OpenAI said it was sharing this information because it believes “it’s important to be transparent with the public and the safety and security communities about this potential shift in capabilities.”

The AI lab said it’s also taking action, including enacting stricter security controls and pausing internal activities involving Astra that don’t meet these beefed guardrails. OpenAI said it is working with relevant government agencies and “select AI safety organizations” to test the capabilities for this model.

When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.

#OpenAI #slowed #Astra #model #development #security #concerns #TechCrunchOpenAI">OpenAI says it slowed Astra model development over security concerns | TechCrunch

OpenAI said Friday it has suspended work on some aspects of its upcoming model Astra after an internal review found it had made significant advancements in agentic coding and cybersecurity — enough to warrant concern over its capabilities.

OpenAI said in a blog post Friday that this model, which is still in development, reached its “critical cybersecurity threshold,” meaning it could independently identify and carry out cyberattacks against traditionally well-protected real-world systems. Under the company’s “Preparedness Framework,” which it created in 2023, this triggered additional safeguards.

“While we continue to benchmark and assess this model, our preliminary evaluations indicate strong enough performance that we cannot rule out Critical capability level at this time,” OpenAI wrote. “Astra is an upcoming model, and was not involved in exploiting Hugging Face.”

The disclosure highlights an unusual moment in the topsy-turvy and still nascent frontier AI labs sector. Companies across every industry hold back products over potential risks, including for safety and cybersecurity concerns. But they rarely announce those decisions publicly when it’s a product that is still under development.

In this case, OpenAI is already under scrutiny after a different unreleased model breached Hugging Face’s systems during internal testing — the first verifiable incident of an AI lab losing control of its model. Since then, OpenAI and AI labs such as Anthropic have disclosed other incidents in which AI models breached their sandboxes and posed threats during cybersecurity tests.

The string of cases — seems like a new disclosure every day now — has triggered varying reactions from cybersecurity experts, lawmakers, and the AI labs themselves. Some express fear and call for stricter oversight. But there’s also a bit of flexing. In certain circles, any AI lab with a model that has that kind of capability will be seen as an impressive advancement.

OpenAI said it was sharing this information because it believes “it’s important to be transparent with the public and the safety and security communities about this potential shift in capabilities.”

The AI lab said it’s also taking action, including enacting stricter security controls and pausing internal activities involving Astra that don’t meet these beefed guardrails. OpenAI said it is working with relevant government agencies and “select AI safety organizations” to test the capabilities for this model.

When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.

#OpenAI #slowed #Astra #model #development #security #concerns #TechCrunchOpenAI

Post Comment