Announcer: This is the AI Law Wiki News Evening Podcast.
Wadilya Saw: Good evening.
Wadilya Saw: It's Friday, October ninth.
Wadilya Saw: Italy sets new safeguards for AI-assisted employment decisions.
Wadilya Saw: Iowa's highest court explains why a speculative deepfake objection failed.
Wadilya Saw: And congressional researchers examine who can be prosecuted when an AI agent goes off course.
Wadilya Saw: I'm Wadilya Saw.
Max Token: And I'm Max Token.
Max Token: We'll also cover government procurement, voiceprint privacy, European AI supervision and automated nuisance calls.
Max Token: This report and the anchor voices were created with generative AI.
Max Token: Our reporting is based on verified sources linked in the show notes.
Wadilya Saw: Italy published a new AI Act implementation decree yesterday, October eighth.
Wadilya Saw: It enters into force October twenty-third.
Wadilya Saw: One of its most consequential provisions concerns people whose jobs are affected by AI-assisted decisions.
Max Token: Article forty-one reserves the final decision to a natural person who exercises effective and autonomous authority.
Max Token: That covers establishing, changing or ending an employment relationship, including disciplinary measures and performance evaluations.
Max Token: The decree says these decisions must not rest solely on automated processing.
Max Token: A worker's consent does not create an exception to that requirement.
Wadilya Saw: The authority language matters.
Wadilya Saw: As we read the provision, the human role must have substance: a person needs actual power over the final outcome.
Wadilya Saw: The text also gives workers a right, on request and through a person, to an understandable explanation of the decision.
Wadilya Saw: That includes AI's contribution and the main parameters considered.
Max Token: There's an important boundary.
Max Token: The decree excludes candidate search and selection from its definition of final decisions establishing employment, even when selection prevents someone from advancing.
Max Token: Listeners should distinguish that stage from the final employment decision protected by this provision.
Wadilya Saw: The decree says employment decisions made in violation of the human-decision requirement are void.
Wadilya Saw: It also requires respect for dignity, privacy and nondiscrimination.
Wadilya Saw: Those protections give the provision consequences beyond a general appeal for responsible AI.
Max Token: Elsewhere, the decree allocates AI market supervision among Italy's national cybersecurity agency and specified financial, insurance and data-protection regulators.
Max Token: The cybersecurity agency becomes the single contact point.
Max Token: For employers and workers, the immediate date to keep in view is October twenty-third, when the decree takes effect.
Wadilya Saw: In Iowa today, the state Supreme Court unanimously affirmed a sexual-abuse conviction and sentence after rejecting a challenge to video evidence.
Wadilya Saw: The defendant argued that the video was inadequately authenticated and might be a deepfake.
Wadilya Saw: We'll focus on the evidence ruling without describing the abuse.
Max Token: Authentication means showing enough evidence to support a finding that an item is what its proponent says it is.
Max Token: Here, the victim could identify private details in the video: the bedroom, furnishings, clothing and the defendant's tattooed hand.
Max Token: Other circumstances also supported the recording's authenticity.
Wadilya Saw: She had been asleep during the depicted conduct.
Wadilya Saw: And investigators had not established a chain of custody connecting the video to the defendant.
Wadilya Saw: The court still found the combined circumstantial evidence sufficient.
Wadilya Saw: It explained that a firsthand witness or testimony about the recording process are not the only ways to authenticate a video under Iowa's rules.
Max Token: A separate issue concerned using a copy rather than the original.
Max Token: Iowa's duplicate-evidence rule allows a duplicate unless a genuine question arises about the original's authenticity, or admitting the duplicate would be unfair.
Max Token: The court found that this defendant offered conjecture about fabrication, rather than a supported reason to doubt this recording.
Wadilya Saw: The opinion recognizes that convincing fabricated videos can incorporate real details.
Wadilya Saw: It also discusses a New York decision that reached a different outcome on a different record.
Wadilya Saw: The practical lesson from today's ruling is that courts examine the foundation and surrounding circumstances of the particular evidence.
Max Token: The court discusses a possible federal rule addressing challenges based on generative-AI fabrication.
Max Token: That remains a proposed rule in this opinion.
Max Token: Today's decision rests on existing Iowa law.
Max Token: It affirms the judgment; the possibility of deepfakes alone did not displace the evidence supporting this video's admission.
Wadilya Saw: An October eighth Congressional Research Service report asks a different accountability question: how federal criminal law applies when an AI agent unexpectedly performs harmful acts.
Max Token: The report says existing laws can reach people who intentionally use AI to commit crimes.
Max Token: Computer hacking, wire fraud and identity theft are among its examples.
Max Token: The harder situation is a person giving lawful instructions to an agent that then takes an unanticipated action.
Wadilya Saw: Many offenses require proof of a particular mental state, such as knowledge or intent.
Wadilya Saw: Rules holding someone responsible for another actor's crime also often depend on intent.
Wadilya Saw: The report explains why those requirements can make prosecution difficult when the human did not anticipate the agent's conduct.
Max Token: For Congress, it describes possible negligence or recklessness standards, developer duties and responsibilities for corporate officers.
Max Token: It also warns of constitutional concerns with criminal punishment imposed without fault.
Max Token: The report analyzes these legislative options.
Wadilya Saw: Congress could instead rely on existing civil remedies, state law, oversight and laws against intentional misuse.
Wadilya Saw: The report's significance is the choice it makes explicit: lawmakers must decide which human conduct and state of mind would justify criminal accountability for an agent's actions.
Max Token: A recently public court order puts the spotlight on what government lawyers said about AI in procurement.
Max Token: The Court of Federal Claims issued the Trax order under seal September eleventh and publicly reissued it October fifth.
Wadilya Saw: Judge Carolyn Lerner found that government declarations inaccurately described who had received AI-generated proposal evaluations.
Wadilya Saw: Recipients included evaluation-board members and the contracting officer.
Wadilya Saw: Earlier representations said that no relevant board member or decisionmaker had viewed those outputs.
Max Token: The government argued that the recipients' roles were limited to evaluating past performance, while the AI material concerned technical evaluation.
Max Token: The court said that distinction did not cure the inaccurate statements.
Max Token: Government counsel acknowledged an error at the hearing.
Wadilya Saw: The judge required four Justice Department and Army lawyers to explain why sanctions should not follow.
Wadilya Saw: Their response was due October fifth; a later response or sanctions outcome has not been verified for this report.
Wadilya Saw: The contract award had already been upheld September third.
Wadilya Saw: The public order concerns candor and record completeness, including material generated or considered during the decision process.
Max Token: Canada's privacy commissioner updated customer-call guidance yesterday.
Max Token: Voiceprints used to authenticate callers are sensitive biometric information, and collecting them may require express consent.
Wadilya Saw: Businesses should tell callers every purpose for recording, clearly and narrowly.
Wadilya Saw: A message about security or quality assurance should not conceal uses for marketing, profiling or authentication.
Wadilya Saw: The guidance distinguishes ordinary implied consent from circumstances involving sensitive information.
Max Token: The duties apply to businesses covered by Canada's federal private-sector privacy law, whether the business or the customer initiates the call.
Max Token: Outsourcing the call centre leaves the business responsible for ensuring the service follows the rules.
Max Token: The guidance also addresses choice, access to recordings, safeguards and retention.
Max Token: It's a practical reminder that a recording can collect identifying information beyond the conversation's words.
Wadilya Saw: The European Commission announced a special scientific-panel meeting for today on frontier-AI safety.
Wadilya Saw: It says the panel has investigated recent loss-of-control incidents and worked with the AI Office on questions for the companies behind the models.
Max Token: The panel's sixty independent experts advise European and national authorities on systemic risks, classification and evaluation.
Max Token: The announcement said recommendations would be presented during the meeting.
Max Token: We have verified the announced activity, rather than a final set of published recommendations.
Wadilya Saw: Separately, the EU's technology chief told Reuters today that the Commission was assessing responses to information requests sent in late August to more than thirty AI companies.
Wadilya Saw: Scientific advice and those supervisory requests are two parts of the EU's response to increasingly capable systems.
Max Token: Britain's privacy regulator, the ICO, says it's actively investigating a surge in automated nuisance calls.
Max Token: Complaints rose from about fourteen hundred in July to nearly thirty-seven hundred in September, the highest level since twenty twenty-one.
Wadilya Saw: Those totals cover automated calls generally.
Wadilya Saw: The regulator also warns about AI-enabled marketing and calls that steer people into messaging groups, personal-data disclosures or fraudulent schemes.
Wadilya Saw: It urges recipients to report suspicious calls through its online tool, because reports help investigators identify patterns and enforcement targets.
Max Token: Finally, Representatives Troy Carter and Bennie Thompson announced the return of their CISA AI task-force bill yesterday.
Max Token: The official congressional record places its introduction and committee referral on October sixth.
Wadilya Saw: According to Carter's announcement, the proposal would create a task force within the Cybersecurity and Infrastructure Security Agency within one year of enactment.
Wadilya Saw: It would coordinate AI security activities, examine workforce gaps and advise stakeholders about deployment risks.
Wadilya Saw: It also calls for twice-yearly congressional briefings.
Wadilya Saw: The bill is pending before the House Homeland Security Committee.
Max Token: That's our Friday report.
Max Token: Italy's October twenty-third commencement is one date ahead; the other developments leave courts, regulators and legislators continuing their work on evidence, privacy and accountability.
Wadilya Saw: Find the full October ninth edition, original sources, transcript and chapters in our show notes.
Wadilya Saw: This has been AI Law Wiki News.
Wadilya Saw: Thanks for listening.