All posts tagged: Safety

UK takes bold measures to enhance AI security for national safety

UK takes bold measures to enhance AI security for national safety

Safeguarding Britain’s national security and protecting citizens from crime will become founding principles of the UK’s approach to AI security from today. Speaking at the Munich Security Conference and just days after the conclusion of the AI Action Summit in Paris, Peter Kyle has today recast the AI Safety Institute the ‘AI Security Institute’. This new name will reflect its focus on serious AI risks with security implications, such as how the technology can be used to develop chemical and biological weapons and how it can be used to carry out cyber-attacks and enable crimes. The Institute will also partner across government, including with the Defence Science and Technology Laboratory, the Ministry of Defence’s science and technology organisation, to assess the risks posed by frontier AI. New approaches to tackle the criminal use of AI As part of this update, the Institute will also launch a new criminal misuse team which will work jointly with the Home Office to conduct research on a range of crime and AI security issues which threaten to harm British …

America’s Deadliest Airliner Crash in 23 Years Happens a Week After Trump Guts Airline Safety Groups

America’s Deadliest Airliner Crash in 23 Years Happens a Week After Trump Guts Airline Safety Groups

Just over a week after firing the heads of the Transportation Security Administration and Coast Guard, eliminating all members of a key aviation security advisory group, and freezing all hiring at the Federal Aviation Administration including key air traffic controllers, the United States has experienced its deadliest aviation disaster in nearly 24 years. Officials fear that all 60 passengers and four crew members on board an American Airlines flight have died after the jet collided with an Army helicopter on a training exercise while landing at Ronald Reagan National Airport near Washington, DC early Wednesday evening. While the timing of the Trump administration’s gutting of aviation safety prior to the crash is impossible to ignore, it’s still unclear what exactly led to the mid-air collision. We know that it was a clear night above the Potomac River, and both aircraft were flying in standard patterns, as the Associated Press reports. Yet obviously something went tragically wrong, and officials believe the crash was preventable. In the immediate wake of the disaster, Trump posted a chilling update …

Parliament Bar To Close For Safety Review Following Alleged Spiking

Parliament Bar To Close For Safety Review Following Alleged Spiking

Houses of Parliament, Westminster, London, Sunset 1 min read17 January A parliament bar has been closed for a safety review following an alleged spiking.  On Friday, a House of Commons spokesperson said: “Strangers’ Bar will close from Monday 20 January while security and safety arrangements are reviewed. “The safety of everyone on the estate remains a key priority of both Houses.” The closure comes after the Metropolitan Police confirmed that they were investigating after a woman reported having her drink spiked in the bar. The incident, which was first reported by Politico, is reported to have taken place on 7 January.   The Stranger’s Bar is used by MPs, parliamentary staff and journalists. MPs are sometimes allowed to take guests. Spiking occurs when an individual gives someone alcohol or drugs without them knowing or agreeing, for example through a drink or even a needle.  Speaking earlier this week, Mike Clancy, General Secretary of Prospect trade union, said: “This is a very concerning alleged incident, especially given the long history of allegations about behaviour in parliament.  “We will be seeking more information about this case …

OpenAI trained o1 and o3 to ‘think’ about its safety policy

OpenAI trained o1 and o3 to ‘think’ about its safety policy

OpenAI announced a new family of AI reasoning models on Friday, o3, which the startup claims to be more advanced than o1 or anything else it’s released. These improvements appear to have come from scaling test-time compute, something we wrote about last month, but OpenAI also says it used a new safety paradigm to train its o-series of models. On Friday, OpenAI released new research on “deliberative alignment,” outlining the company’s latest way to ensure AI reasoning models stay aligned with the values of their human developers. The startup used this method to make o1 and o3 “think” about OpenAI’s safety policy during inference, the phase after a user presses enter on their prompt. This method improved o1’s overall alignment to the company’s safety principles, according to OpenAI’s research. This means deliberative alignment decreased the rate at which o1 answered “unsafe” questions – at least ones deemed unsafe by OpenAI – while improving its ability to answer benign ones. Graph measuring o1’s improved alignment compared to Claude, Gemini, and GPT-4o (Image Credit: OpenAI) As AI …

Embattled Character.AI Hiring Trust and Safety Staff

Embattled Character.AI Hiring Trust and Safety Staff

Content warning: this story discusses sexual abuse, self-harm, suicide, eating disorders and other disturbing topics. Character.AI, the Google-backed AI chatbot startup embroiled in two lawsuits concerning the welfare of minors, appears to be bulking up its content moderation team in the wake of litigation and heightened public scrutiny. The embattled AI firm’s trust and safety head Jerry Ruoti announced in a LinkedIn post yesterday that Character.AI is “looking to grow” its safety operations, describing the role as a “great opportunity” to “help build a function.” A linked job listing for a “trust and safety associate,” also posted yesterday, describes a role akin to a traditional social media moderation position. Contract hirees will be tasked to “review and analyze” flagged content for “compliance with company moderation standards,” remove content deemed “inappropriate or offensive,” and “respond to user inquiries” concerning safety and privacy, among other duties. The apparent effort to manually bolster safety teams comes as Character.AI faces down two separate lawsuits filed on behalf of three families across Florida and Texas who claim their children were …

Common chemical in drinking water hasn’t been tested for safety

Common chemical in drinking water hasn’t been tested for safety

Millions of US residents may be drinking water containing the potentially harmful compound Yiu Yu Hoi/Getty Images A common disinfectant in drinking water breaks down into a chemical compound that we know almost nothing about, including whether it has any potential toxic health effects to those who drink it. Chlorine has been used to sanitise drinking water for more than a century. However, some drinking water systems in the US, UK and Australia now use another closely related chemical disinfectant called chloramine. That’s because chlorine byproducts were linked to bladder and colon cancer, low birth rates and miscarriage, says … Source link

Instagram unveils new feature as govt tightens online safety rules | Science, Climate & Tech News

Instagram unveils new feature as govt tightens online safety rules | Science, Climate & Tech News

Instagram is releasing a feature that will let users easily reset their algorithms, as the government strengthens its regulation of online safety. With the new reset feature, users can clear their recommended content from Explore, Reels and their feed, potentially reducing the amount of harmful content they are exposed to. It’s all part of Meta’s push to make the app safer for young people, after announcing more private Teen accounts in September. The feature, which will soon be rolled out globally, was announced as the government outlined its priorities for online safety. Peter Kyle, Labour’s technology secretary, said Ofcom should ensure the concept of “safety by design” is being followed by tech companies from the outset. That would ensure more harm is caught before it occurs. Image: Peter Kyle (L). Pic: Reuters He also pushed for more transparency from tech giants on what harms are occurring on their platforms. “From baking safety into social media sites from the outset, to increasing platform transparency, these priorities will allow us to monitor progress, collate evidence, innovate, and …

California’s governor blocked landmark AI safety laws. Here’s why it’s such a key ruling for the future of AI worldwide

California’s governor blocked landmark AI safety laws. Here’s why it’s such a key ruling for the future of AI worldwide

In a world where artificial intelligence is rapidly shaping the future, California has found itself at a critical juncture. The US state’s governor, Gavin Newsom, recently blocked a key AI safety bill aimed at tightening regulations on generative AI development. The Safe and Secure Innovation for Frontier Artificial Intelligence Models Act (SB 1047) was seen by many as a necessary safeguard on the technology’s development. Generative AI covers systems that produce new content in text, video, images and music – often in response to questions, or “prompts”, by a user. But Newsom said the bill risked “curtailing the very innovation that fuels advancement in favour of the public good”. While agreeing the public needs to be protected from threats posed by the technology, he argued that SB 1047 was not “the best approach”. What happens in California is so important because it is the home of Silicon Valley. Of the world’s top 50 AI companies, 32 are currently headquartered within the state. California’s legislature therefore has a unique role in efforts to ensure the safety …

AI safety showdown: Yann LeCun slams California’s SB 1047 as Geoffrey Hinton backs new regulations

AI safety showdown: Yann LeCun slams California’s SB 1047 as Geoffrey Hinton backs new regulations

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Yann LeCun, chief AI scientist at Meta, publicly rebuked supporters of California’s contentious AI safety bill, SB 1047, on Wednesday. His criticism came just one day after Geoffrey Hinton, often referred to as the “godfather of AI,” endorsed the legislation. This stark disagreement between two pioneers in artificial intelligence highlights the deep divisions within the AI community over the future of regulation. California’s legislature has passed SB 1047, which now awaits Governor Gavin Newsom’s signature. The bill has become a lightning rod for debate about AI regulation. It would establish liability for developers of large-scale AI models that cause catastrophic harm if they failed to take appropriate safety measures. The legislation applies only to models costing at least $100 million to train and operating in California, the world’s fifth-largest economy. Most of these signatories have a distorted view of what is coming next with AI.The distortion is due to their inexperience, naïveté on how difficult the …

Grenfell MP Says Building Safety Crisis Must Be Top Of Government Agenda

Grenfell MP Says Building Safety Crisis Must Be Top Of Government Agenda

Seven years since the fire, Grenfell campaigners have criticised the inquiry for taking too long (Alamy) 6 min read01 September Labour MP for Kensington and Bayswater Joe Powell has said that the Grenfell Inquiry report coming out this week will be a “huge moment” for the community, and hopes it will lead to the Government holding a debate on building safety in Parliament. A devastating fire broke out in the Grenfell Tower high-rise block in Kensington, London, on 14 June 2017, killing 72 people and injuring more than 70 others. It was the worst residential fire in the UK since the Blitz of World War II. On Wednesday, Phase 2 of the Grenfell Tower Inquiry will be published, expected to look in more detail at the causes of the fire and the decision-making by authorities and companies which led to it. The report will be made available to bereaved and survivors on Tuesday before going public at 11am on Wednesday, with a Government statement expected later that afternoon. The inquiry was …