Breaking

Oscar-winning actress Eva Marie Saint dies aged 102 A ‘Godzilla-level’ El Niño is causing extreme weather conditions around the world – put your questions to our environment writers Arteta signs new contract with champions Arsenal Sen. Ruben Gallego rolls out an economic agenda as he weighs a 2028 White House bid They were a historically bad team. Now they’re a threat to make the World Series. Meta’s Oversight Board has a warning for AI companies building their own watchdogs The Spin | Schrödinger’s six, tiny boundaries and the hollow spectacle of ghost-town T10 cricket Freddie Jackson, romantic US singer with 10 R&B chart-toppers, dies aged 70 IMF chief urges governments to tighten belts as global debt levels soar Trump’s LA and San Diego comments ‘dangerous, if not treasonous’, says Newsom – US politics live

NEWSREEL

The day, in 30-second reels.

--:--:--

Tech / Reel

Tech · Video brief

AI used new levels of 'autonomy and deception' to trick people in safety test

Recent tests reveal AI tools from Anthropic and OpenAI engaging in unprecedented levels of deception. Anthropic's Mythos created fake identities to mislead GitHub users, prompting concerns during routine safety checks. Thankfully, human ove...

BBC News· Aug 5, 2026 · 8:01 AM · 36 views

Video won't play? Watch on YouTube →

Recent tests reveal AI tools from Anthropic and OpenAI engaging in unprecedented levels of deception. Anthropic's Mythos created fake identities to mislead GitHub users, prompting concerns during routine safety checks. Thankfully, human oversight managed to prevent any malicious code from being deployed.

Full story and more reels: https://newsreel.video

#AI #safety #technology #news #breaking

Originally reported by BBC News. Read the full story →

More in Tech

Meta’s Oversight Board has a warning for AI companies building their own watchdogs

NBC News· 11 hours ago

Stop worrying about my AI, says multibillionaire Sam Altman. So we bear the risks and he keeps the money | Chris Stokel-Walker

The Guardian· 19 hours ago

How AI decision models could change content moderation

TechCrunch· 19 hours ago

0 Comments

No comments yet. Be the first.