live from washington, this is bbc news. billions head to the polls around the world this year, as concerns grow about the impact of deceptive artificial intelligence on elections. us officials warn foreign adversaries could use the technology to influence elections, as the deep fakes become more convincing and sophisticated. hello, i'm katrina perry. you're very welcome to this bbc special report on the influence of artificial intelligence on global elections. more than 60 countries will head to the polls in 2024, and many already have. half the world's voting population, about 2 billion people are expected to cast ballots. artificial intelligence has

already begun to play substantial role in elections. and concerns are growing about how to protect the public from deepfakes that could sway their vote, especially as the technology becomes more sophisticated and convincing. in pakistan, an ai version ofjailed imran khan was released by his own team, claiming victory in feburary. and in argentina's elections, the new york times wrote an article showing multiple images released by both campaigns were digitally doctored. here in the us, voters in new hampshire recieved a fake robocall from presidentjoe biden — telling them not to vote in the state's primary back in january. take a listen. what a bunch of malarkey. we know the value of voting democratic, when our votes count, it is important you save your vote for the november election. we will need your help in electing democrats up and down the ticket. voting this tuesday only enables the

republicans in their quest to elect donald.— republicans in their quest to elect donald. obviously not joe biden there, _ elect donald. obviously not joe biden there, but _ elect donald. obviously not joe biden there, but was _ elect donald. obviously not joe biden there, but was an - elect donald. obviously not joe biden there, but was an ai - biden there, but was an ai generated deepfake. however, tech giants are pushing back. in feburary, 20 major companies — including adobe, amazon, google, ibm, meta, microsoft, 0penai, tiktok, and x — signed the tech accord at the munich security conference, an agreement to join forces to combat deepfakes and protect elections. more recently,tiktok annnounced on thursday that it will begin labeling ai—generated content. meanwhile, meta said last month it will do the same. the google—owned video—sharing platform youtube says it requires creators to disclose when realistic content is ai—generated. this week microsoft and open ai announced a $2 million societal resilience fund to further ai literacy. just before coming on air i spoke to ginny badanes, head of microsoft's democracy forward initiative, about how tech companies can protect against bad actors.

thank you forjoining us on bbc news. if we look out, as we are in this year, so many elections, if we look at the issue of deep steaks and ai generated content, some of it is very hard to tell but it is not true. how can people know what they are listening to, what they are listening to, what they are seeing, is authentic? it what they are seeing, is authentic?— authentic? it is a great question- _ authentic? it is a great question. this - authentic? it is a great question. this is - authentic? it is a great question. this is a - authentic? it is a great| question. this is a huge authentic? it is a great - question. this is a huge year for elections, something like 2 billion people are going to the polls to vote in consequential elections. when generative ai is on the rise at the same time it is important people have tools they can use to decipher what ai is, what they can trust, and they also need to be able to access authoritative information. can they look at an image and know right away if it is a real or not? not necessarily, not yet, but there are some indicators if it is telling your story that is maybe not true, you can go look

for the original source of the information to see if you can get additional context around what that puts a lot of emphasis tech companies be doing more?— emphasis tech companies be doing more? absolutely. there is a big role _ doing more? absolutely. there is a big role for— doing more? absolutely. there is a big role for the _ doing more? absolutely. there is a big role for the tech - is a big role for the tech companies in this. there are lots of promises around indicators of trust, labelling and watermarking that the companies are investing in and which they are starting to make more available and visible to the end user. the reality is where that technology is unaware labelling is now, we also need to invest in al literacy and ai training, so that people on the other side know what questions they should be asking. at your department to track and chart how much of this is out there. how much of this is out there. how much of this material a finding, especially that aimed at destabilising democracies or interfering with elections? we have been _ interfering with elections? - have been tracking especially what nationstates are doing in this space. there have been so many elections we can look at so far and see what has happened there stop so far we have not seen massive amounts

of ai used to deceive people around elections, but that doesn't mean it isn't coming and it doesn't mean others aren't planning for that. it is important for us to continue reporting on what we are seeing, what we're seeing these adverse areas do, and equipping people, especially political campaigns and election officials, with what they should do if they encounter about. . , ., should do if they encounter about. . ., about. can you say your own microsoft — about. can you say your own microsoft products _ about. can you say your own microsoft products are - about. can you say your own microsoft products are or. about. can you say your own | microsoft products are or are not being used to because harm and contribute to this? we not being used to because harm and contribute to this?- and contribute to this? we are doinu and contribute to this? we are doin: the and contribute to this? we are doing the best _ and contribute to this? we are doing the best we _ and contribute to this? we are doing the best we can - and contribute to this? we are doing the best we can to - doing the best we can to hopefully not have that be the case. we have different policies, restrictions on our products, to ensure they are not being used for those purposes. and we have policies and enforcement if we determine that somebody has used them in a deceitful way or a way that goes around what our policies state. if goes around what our policies state. ., goes around what our policies state. . . ., state. if an election campaign, if a candidate _ state. if an election campaign, if a candidate in _ state. if an election campaign, if a candidate in any _ state. if an election campaign, if a candidate in any country, l if a candidate in any country, i am not talking specifically about the us, but if they want to target their opponents with deceptive ai and they want to share that to their own existing mailing list of supporters, is there anything any tech company can do about that? if they are using it in

that? if they are using it in that way and we are able to determine they have done so, as the provider of the technology, we can cut off usage for sure. so there are some ways to stop it in the first place. there are also tools to prevent that kind of use in the beginning. i think what we are looking for is less likely to be campaigns using it for this purpose, and more likely actors outside the political space looking to influence elections. that is where we are looking at our press intelligence team, which can be really helpful, to see what these actors are doing, how they are trying to manipulate the public, sharing that information with governments and with the public to know what to look out for. are there particular things being targeted? so are there particular things being targeted?— are there particular things being targeted? so far we see russia, being targeted? so far we see russia. for — being targeted? so far we see russia, for example, - being targeted? so far we see russia, for example, is - russia, for example, is continuing to put propaganda about ukraine out into the space, including places where elections are coming up in order to potentially influence those people around votes. hard to say for sure what their intentions are. we have seen different techniques like deep fakes being used. —— deepfakes. they are not particularly effective at the moment, these

technologies are improving in these actors are looking to leverage the best technology out there. so it could continue over time. out there. so it could continue over time-— out there. so it could continue over time. microsoft and about 20 other companies _ over time. microsoft and about 20 other companies have - 20 other companies have obviously signed the tech accords, but that is self policing, self—regulating. there is no legislation in this country. the eu has its ai act. would you like to see legislation? should there be legislation? should there be legislation?— legislation? should there be legislation? there should be. the tech accord _ legislation? there should be. the tech accord was - legislation? there should be. the tech accord was a - legislation? there should be. i the tech accord was a moment legislation? there should be. - the tech accord was a moment in time when the tech companies came forward to acknowledge the challenges we are facing in this election cycle and the reality of the timeline, which is that it takes a while to create that legislation and enforce it. but we are already in the middle of an election cycle. the voluntary commitments were the tech industry's way of saying, we are going to step into this gap, we are identifying these challenges in committing to what we can do about it. that is not in lieu of regulation. we absolutely want to see additional regulation from the us and other governments. what us and other governments. what about freedom _ us and other governments. what about freedom of _ us and other governments. what about freedom of expression? there has been political satire since the beginning of time and some would say some of this ai generated stuff is just that.

al generated material is not ai generated material is not necessarily bad. in fact, political campaigns using ai material is not necessarily bad. it is when it comes with the intent to deceive we're drawing a line. if you put your words and else's mouth, that is fraud, and that is definitely crossing a line. if you use ai crossing a line. if you use al to make a funny video about yourself, that is absolutely acceptable from our perspective. so there is a difference between using the technology to further a message and using it to deceive others. 0k. and using it to deceive others. ok. we will leave it there for the moment. thanks forjoining us on bbc news. i spoke to adam presser, tiktok�*s head of operations and trust and safety. i asked him what they are doing specifically to detect and stop deceptive actors from targeting online platforms when it comes to elections. keeping our communities safe is our top priority, especially through these really important processes. we have navigated it

over 150 elections, over the course of the last several years, all over the world, so we work with partners like democracy works to provide reliable information to people searching for information about the election in the united states, for example. we also have 18 fact checking partners to ensure the content people are seeing is reliable and has been verified. so it really requires that multipronged approach to ensure the experience on tiktok is as safe as possible. in experience on tiktok is as safe as possible-— as possible. in the us, tackling _ as possible. in the us, tackling the _ as possible. in the us, tackling the threat - as possible. in the us, tackling the threat of l tackling the threat of artificial intelligence before the november presidential election has become a significant priority across the government. the fbi has warned that foreign adversaries could use ai that foreign adversaries could use al to interfere in the election and spread disinformation. 0ne senior official says it is a concern that we will probably see growth and in the coming years. the department of homeland security has also want election officials and generative ai could be used to enhance foreign influenced campaigns targeting the 2024 cycle. and

in congress, lawmakers are discussing legislation on al's impact in federal elections, but it is unlikely any will pass by november. 0ne but it is unlikely any will pass by november. one of those proposals is the protect elections from deceptive ai act, a bill that would ban the intentional publishing of materially deceptive ai generated political ads which are intended to influence an election or indeed to raise money. earlier i spoke to one of the bill's authors, democratic senator chris coons from delaware.— from delaware. you are not facin: from delaware. you are not facing the _ from delaware. you are not facing the electorate - from delaware. you are not i facing the electorate yourself this year, but plenty of your democratic colleagues are, including presidentjoe biden, a close confidant of yours. haifa a close confidant of yours. how concerned _ a close confidant of yours. how concerned is _ a close confidant of yours. how concerned is he, _ a close confidant of yours. how concerned is he, is _ a close confidant of yours. how concerned is he, is the - concerned is he, is the campaign, is the party as a whole, about the role that deceptive ai may play in this year's election?— year's election? well, thank ou for year's election? well, thank you for a — year's election? well, thank you for a chance _ year's election? well, thank you for a chance to - year's election? well, thank you for a chance to be - year's election? well, thank you for a chance to be on i year's election? well, thankl you for a chance to be on and thanks for your focus you for a chance to be on and thanks for yourfocus on you for a chance to be on and thanks for your focus on this important subject. i am very concerned about the rapidly growing capabilities of ai

concerned about the rapidly growing capabilities of al to carry out, for example, fake telephone calls mimicking somebody�*s voice. we already saw an episode of this in the new hampshire primary here. we saw deepfakes used to try to influence the election in taiwan. we have seen it have an impact on the outcome of the election in slovakia. so there is already electioneering a foot, where ai fakes, whether they are video or audio, can and i think increasingly will be used to try to deceive the public, to dissuade them from voting, orto public, to dissuade them from voting, or to spread lies about what different candidates for office are doing or what they stand for. office are doing or what they stand for-— office are doing or what they stand for. senator, have you seen any evidence _ stand for. senator, have you seen any evidence of- stand for. senator, have you seen any evidence of that i seen any evidence of that coming from foreign actors in this country?— coming from foreign actors in this country? yes. would you like to elaborate? _ this country? yes. would you like to elaborate? well, - this country? yes. would you | like to elaborate? well, things that i like to elaborate? well, things that i know _ like to elaborate? well, things that i know from _ like to elaborate? well, things that i know from classified - that i know from classified briefings i am not going to talk about in detail. i will just say that the previous segment, where the person being interviewed was saying that a number of active foreign

players had been using artificial intelligence to strengthen their disinformation campaigns with an intention to influence our elections, that is quite accurate, and a testimony, the director of the fbi has said publicly, we have a number of adversaries, whether it is russia, china, iran or the dprk, north korea, that are using disinformation and now artificial intelligence to seek to interfere with our elections. to seek to interfere with our elections-— to seek to interfere with our elections. ., ., ., elections. you are a co-sponsor ofthe elections. you are a co-sponsor of the act, _ elections. you are a co-sponsor of the act, it — elections. you are a co-sponsor of the act, it has _ elections. you are a co-sponsor of the act, it has broad - of the act, it has broad bipartisan support, one of the few issues in this country that there is agreement on. why can that not be rushed through? why are we unlikely to see that in action before november to protect against what you are describing there? i protect against what you are describing there?— describing there? i was encouraged _ describing there? i was encouraged that - describing there? i was encouraged that the i describing there? i was i encouraged that the rules committee, which senator clover charge chairs, has been scheduled for a markup this coming week. —— senator klobuchar. frankly we have struggled to enact reasonable, responsible, meaningful

protections in privacy generally. in data regulation more broadly. and in artificial intelligence. senators klobuchar and holly are leading this bill and doing a good job, and if there is one piece of legislation i would like to see past before our elections, it would be this one. do past before our elections, it would be this one.— past before our elections, it would be this one. do you think it mitht? would be this one. do you think it might? it _ would be this one. do you think it might? it might. _ would be this one. do you think it might? it might. we - would be this one. do you think it might? it might. we are - it might? it might. we are havin: it might? it might. we are having difficulty _ it might? it might. we are having difficulty getting i it might? it might. we are i having difficulty getting along in congress and moving things through the senate and the house, butjust the very breath of radiological spread between senator hawley, who was on the far right, and senator klobuchar, who like me is a democrat, gives you some indication of how much support there might be for this. i had a hearing last week on another bipartisan artificial intelligence related bill that seeks to create a private right of action, a legal protection, for anybody, whether they are a politician or an artist, an actor orjust an average actor or just an average citizen, actor orjust an average citizen, who is the victim of a deep faith or an ai avatar

being used to make them say or do things they never approved of or don't agree with. $5 do things they never approved of or don't agree with.- of or don't agree with. as we heard earlier, _ of or don't agree with. as we heard earlier, deepfakes - of or don't agree with. as we heard earlier, deepfakes are | heard earlier, deepfakes are being used when it comes to conflicts as well, in propaganda and so on. i wanted to ask you about another issue, some news developing tonight, this state department report on israel's use of american supplied weapons. this report says it is reasonable to assess they have been used in gaza in breach of international humanitarian law by israel. the report stops short of recommending the us should withhold weapons as a result, but how does the united states move forward with israel now if weapons are being used, most likely in this way? $5 weapons are being used, most likely in this way?— likely in this way? as you know, president - likely in this way? as you know, president biden i likely in this way? as you | know, president biden has likely in this way? as you - know, president biden has made clear, first privately and more recently publicly, that he doesn't support a full—scale invasion of rafah to go after hamas until civilians are allowed to move out of the way. that is something he and prime minister netanyahu have minister neta nyahu have discussed minister netanyahu have discussed many times, and i

think the concerns raised by our president and by many in congress about balancing israel's right, even its obligation, to protect its citizens from hamas, by continuing to go after hamas in rafah, needs to be balanced with obligations under international humanitarian law to protect civilians and allow them to move out of the way. i think that is whatjust one application of the broader insights you are describing as having come out of this report that was released apparently late today. that was released apparently late today-— late today. just very briefly, senator. — late today. just very briefly, senator. is _ late today. just very briefly, senator, is it _ late today. just very briefly, senator, is it concerning - late today. just very briefly, senator, is it concerning to | senator, is it concerning to you that us weapons would use intentionally in breach of international humanitarian law? does that concern you? fit does that concern you? of course. — does that concern you? of course. it _ does that concern you? of course, it should concern any of us if weapons we have provided to many other countries are used in a way that violates our values and principles, but i think the important thing here is that our close and vital ally israel be allowed to defend itself. we came to their aid and defence when they were attacked by 300

iranian drones and missiles just a few weeks ago, that they be armed and able to attack air and prevent them from carrying out any more attacks, but that they do so in a way that meets they do so in a way that meets the test of international humanitarian law. shill the test of international humanitarian law. all right, we will leave it _ humanitarian law. all right, we will leave it there _ humanitarian law. all right, we will leave it there for _ humanitarian law. all right, we will leave it there for a - will leave it there for a moment. senator chris combs, democrat delaware, thank for joining us. democrat delaware, thank for joining us— joining us. outside of federal legislation. _ joining us. outside of federal legislation, many _ joining us. outside of federal legislation, many us - joining us. outside of federal legislation, many us states | joining us. outside of federal. legislation, many us states are looking for another way to tackle the problems of ai. nearly all of the state legislatures currently in session are considering ai—related bills, and nearly half of those bills address deepfakes. that's according to data shared with axios. as of february 7, there were 407 total ai—related bills across more than 40 us states, that's up from just 67 bills last year. legislators in 33 states have put forward specific election—related ai bills. this includes legislation co—sponsored by michigan state representative matthew bierlein. representative bierlein's law went into effect in michigan in february and it regulates and establishes restrictions on the use of ai in

political campaigns. i spoke to him earlier. the biggest concern isjust the biggest concern is just how easy it is. you don't have to be a tech genius any more to access the technology. to make the adjustments to a speech, or images, and then put it out there as a legitimate situation with a candidate. and so we just thought the best thing we could do was pass some slate —— state legislation and have as much control of our process as we can. i5 much control of our process as we can. . . much control of our process as we can. , ., ., , ., we can. is there a great use of ai we can. is there a great use of ai generated _ we can. is there a great use of ai generated content - we can. is there a great use of ai generated content there? l we can. is there a great use of| ai generated content there? is it something report —— that voters are reporting to you which is a matter of concern? we haven't heard yet, i have not heard yet of any of those things. wejust not heard yet of any of those things. we just know that it is so simple, the deep fake technology, it isjust so simple, the deep fake technology, it is just a readily available, and the way we thought would be the best way to do that, from our

election standpoint, we're not telling you can't use it. we are not telling you that you can't make things up about your opponent. what we are telling you is that if you are going to use it in campaign related material, you have to put a disclaimer on it saying it is al generated, and that if it ai generated, and that if it does not have that it is not going to be valid campaign material. . going to be valid campaign material-— material. have you given consideration _ material. have you given consideration to - material. have you given consideration to telling l material. have you given - consideration to telling people they cannot make stuff up about their opponent, as you put it there? i their opponent, as you put it there? .. . their opponent, as you put it there? ~' , there? i think there is some olitical there? i think there is some political licensing _ there? i think there is some political licensing every - political licensing every election, in every campaign, the problem that we have with al material is that it is so ai material is that it is so advanced, it isjust ai material is that it is so advanced, it is just not reasonable for a person to think that is made up. in the past, when you have sent out a piece of mail showing your opponent, you know, robbing the taxpayers, holding bags of money, everybody knows that that was a photoshopped image that was a photoshopped image that didn't really happen. that is not what we're talking about anymore. the phone call that

was mentioned earlier, when we had our committee meeting, where we discussed this legislation, we played a video, monitor video, sorry, we played a speech of president biden telling the committee that they to pass this legislation that was so crucial, and one of my co—sponsors made that speech up and just plugged in the name they wanted to read it. it took them about four minutes. those other kinds of things are dealing with now.- other kinds of things are dealing with now. and the campaigns. _ dealing with now. and the campaigns, their - dealing with now. and the i campaigns, their supporters, the public action committees, are they obeying the law? 50 are they obeying the law? so far we are they obeying the law? sr far we have not had any complaints. the way that people could address it is by campaign finance violation, basically, with our secretary of state, or they can seek an injunction through the attorney general�*s office. we have not yet had any reports of that, but we are really just getting started with the campaign season, first primary elections in august. we

have been _ primary elections in august. we have been talking about the negative impacts of ai and election cycles, is that a place for the positive use of it, do you think?— place for the positive use of it, do you think? with any new technology. — it, do you think? with any new technology. we _ it, do you think? with any new technology, we always - it, do you think? with any new technology, we always have i it, do you think? with any new technology, we always have to weigh up the positives and the negatives. i am sure there are good things out there. we use ai, when we are trying to figure out complex maths problems on things like that. we use it when we are designing new cars and houses, to help fill in the gaps, for something that the human operatorjust missed out first. but when it comes to campaign material, you should run on your own merits, is my opinion. should run on your own merits, is my opinion-— is my opinion. that is your rood is my opinion. that is your good advice _ is my opinion. that is your good advice there. - is my opinion. that is your good advice there. we - is my opinion. that is yourj good advice there. we will leave it there for the moment. michigan state representative matthew billy, thanks for joining us. matthew billy, thanks for joining us— matthew billy, thanks for “oinin us. . ., . joining us. thanks for having me. it's a different story across the pond as the european union successfully passed legislation combatting deceptive ai. in march, eu lawmakers approved the world's first comprehensive framework for constraining the risks of artificial intelligence.

the ai act works by classifying products according to risk and adjusting scrutiny accordingly. the law's creators said it would make the tech more "human—centric." the law doesn't come into effect until 2025. last month, politico exclusively reported that chatbots produced by google, microsoft and 0penai shared some false information about the european election. citing research done by the berlin—based ngo democracy reporting international, researchers found that the chatbots returned incorrect election dates and information about how to cast a ballot. we heard from jenny earlier in the programme about how they had not seen massive amounts of deceptive ai being used around election. senator chris combs said he has classified information there is some use of it in this country. how serious a problem do you think it is? i think there is a broader problem in the information space widely, in the sense that we have had misinformation and disinformation for centuries. now we have a digital space where information is, to use

the word of the year, i guess, turbocharged, and people can get a variety of information from a lot of different places. there has been a fracturing of the online space, content moderation challenge as well. there is a big back challenge for the information space. and when you add another layer to that, which is the ai generated potential, i think thatjenny is correct, in a sense, that of the content that we are seeing, we can think about israel and gaza and the initial conflict, the flooding of the information space when supply was low but demand is extremely high, much of it wasjust demand is extremely high, much of it was just recycled images, decontextualised clips, things like that. it was not really ai generated content. and it very well could have been, and within the sort of framework that we are operating, should have been, in some respects. i think ai generated content is a layer added onto this, that amplifies a lot of what we see, thatis amplifies a lot of what we see, that is a new tool, and then of

course there are some unique challenges to the ai mode as well, especially for the information space. i would say that broadly it is part of a bigger challenge for information.- bigger challenge for information. �* ., information. and if foreign actors want _ information. and if foreign actors want to _ information. and if foreign actors want to destabilise | actors want to destabilise democracy and interfere in elections, they have always been able to do that, this just makes it easier. been able to do that, this 'ust makes it easier.i makes it easier. yes, so thinking _ makes it easier. yes, so thinking about _ makes it easier. yes, so i thinking about spearfishing attacks, right? —— spear—phishing. there is clunky language in spear—phishing attacks, the social engineering was awkward in terms of trying to get you to download a file from your maybe colleague or something like that. now it is potentially going to be tailored to you. because they can pull information about you or people like you, to be a little bit more targeted. there are open models being sold on the dark web which actually write malware, lowering the threshold for writing, some of this up cyber security

challenges, i think that is one element that is going to be amplifying the number of actors, potentially, in this space, especially as we think about cyber security as typically the purview of nationstates, now they can be many more actors in play. so thatis many more actors in play. so that is one element, the social engineering side, and of course the other element is the deep fake side, which i think the senator spoke about. it is really, especially audio, there is something really hard about processing audio, and of course thatjoe biden robo call was discovered immediately, but there are many elections, the national elections can hinge on local level contests, or local level concerns.— local level contests, or local level concerns. smaller numbers of votes. level concerns. smaller numbers of votes- so. _ level concerns. smaller numbers of votes. so, what _ level concerns. smaller numbers of votes. so, what can _ level concerns. smaller numbers of votes. so, what can users, - of votes. so, what can users, consumers and voters actually do? it seems like a lot of it is on the individual to try to weed out for themselves whatever the laws may be, whatever the laws may be, whatever the laws may be, whatever the tech companies may say. i whatever the tech companies may sa . ~' whatever the tech companies may sa . ~ ., , say. i think with anything in this space _

say. i think with anything in this space there _ say. i think with anything in this space there is - say. i think with anything in this space there is a - say. i think with anything in i this space there is a challenge of reactivity. so right now, and i thinkjenny also mentioned this as well, while the space for the tech is very good, it is still a bit glitchy. it has some bugs. the very high quality content, stuff that is coming out of open ai and other companies, they are adding in layers of protection, and so where the home is going to happen, it might be there, but they are working hard to put in safeguards, and other tools that i may be less sophisticated will still have candidates donald trump of seven fingers or something odd going on in the back ground, potentially. so being aware of some of these things is really important, still, buti some of these things is really important, still, but i also think it shifts the onus onto the viewers, to pass out some of those challenges. bind the viewers, to pass out some of those challenges. and those companies _ of those challenges. and those companies are _ of those challenges. and those companies are of— of those challenges. and those companies are of course - of those challenges. and those companies are of course each | companies are of course each day trying to do things to help their users along, aren't they. thank you so much forjoining us. those insights are

excellent. that is it for the moment. i will be back shortly with the news headlines. stay with the news headlines. stay with us here on bbc news. hello there. settled and warm again on friday, with plenty of late spring sunshine around and temperatures rising across the four nations into the low 20s in celsius. the warmth is set to last as we head through the weekend. temperatures will remain above the seasonal average, warmest towards the east. and it's still dry for the vast majority of us on saturday. a scattering of showers, with the real breakdown happening on sunday. heavy, thundery showers out towards the western half of the uk. further east should stay largely dry. and here is the area of high pressure that's keeping these dry, settled conditions for the time being. it will eventually push further eastwards into scandinavia. but we've got a bit of an easterly breeze, and that's been dragging some mist and low cloud in from the north sea. but that will lift and clear across the south—east of england and east anglia through saturday morning. still maybe a hang—back of cloud towards parts

of the yorkshire lincolnshire coast, though. lots of sunshine to start the day and we'll keep the sunny skies for most through the afternoon. but a scattering of showers across scotland pushing northwards, perhaps some heavy and thundery, but they'll be fairly isolated. it's still very warm — 24 degrees celsius in glasgow. chance of a shower, too, across northern areas of northern ireland and north wales. a little cooler towards these north sea—facing coasts, with some of the cloud possibly lapping onshore again at times. 25 or 26 degrees celsius in london and south—east england. so the high pressure starts to push further eastwards as we head through sunday. that allows for these low pressure systems to roll in from the west. and this weather front will bring us thickening cloud across the south—west of england, western wales, on sunday morning. some showers across the western isles and western scotland, pushing into northern ireland, and the chance of some thunderstorms developing all across the western half

of the uk. but it should stay drier further east. again, there will be a lot of sunshine here, and once again we could see temperatures in the low to the mid—20s in celsius. but cooler out towards the west, of course, underneath the cloud and with the eventual rain. and here comes that low pressure system swinging in as we head through monday. it's going to give us quite widespread rain on monday, especially through the afternoon, so expect it to turn a lot more showery as we head through next week. and there'll be a drop in temperature, too, so unsettled and cooler as we head through next week. bye—bye for now.

voice-over: this is bbc news. we'll have the headlines for you at the top of the hour, which is straight after this programme. i'm in the presidential palace in buenos aires to interview the new president, javier milei. javier milei went viral during his campaign around the world for campaigning with a chainsaw, which was his symbol for his desire to slash the state and slash public spending. after years of high public spending, high debt, high inflation, he has slashed the value of the currency, slashed subsidies, slashed government departments. he says this radical programme of austerity is needed to fix the economy, but lots of people in argentina are struggling, and he's said that things will need to get worse before they get better. he's called himself a libertarian and anarcho—capitalist. he hates the state,

he loves elon musk. he's a big fan of the rolling stones.